xref: /openbmc/linux/kernel/signal.c (revision 3aa139aa9fdc138a84243dc49dc18d9b40e1c6e4)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  *  linux/kernel/signal.c
4  *
5  *  Copyright (C) 1991, 1992  Linus Torvalds
6  *
7  *  1997-11-02  Modified for POSIX.1b signals by Richard Henderson
8  *
9  *  2003-06-02  Jim Houston - Concurrent Computer Corp.
10  *		Changes to use preallocated sigqueue structures
11  *		to allow signals to be sent reliably.
12  */
13 
14 #include <linux/slab.h>
15 #include <linux/export.h>
16 #include <linux/init.h>
17 #include <linux/sched/mm.h>
18 #include <linux/sched/user.h>
19 #include <linux/sched/debug.h>
20 #include <linux/sched/task.h>
21 #include <linux/sched/task_stack.h>
22 #include <linux/sched/cputime.h>
23 #include <linux/file.h>
24 #include <linux/fs.h>
25 #include <linux/proc_fs.h>
26 #include <linux/tty.h>
27 #include <linux/binfmts.h>
28 #include <linux/coredump.h>
29 #include <linux/security.h>
30 #include <linux/syscalls.h>
31 #include <linux/ptrace.h>
32 #include <linux/signal.h>
33 #include <linux/signalfd.h>
34 #include <linux/ratelimit.h>
35 #include <linux/tracehook.h>
36 #include <linux/capability.h>
37 #include <linux/freezer.h>
38 #include <linux/pid_namespace.h>
39 #include <linux/nsproxy.h>
40 #include <linux/user_namespace.h>
41 #include <linux/uprobes.h>
42 #include <linux/compat.h>
43 #include <linux/cn_proc.h>
44 #include <linux/compiler.h>
45 #include <linux/posix-timers.h>
46 #include <linux/cgroup.h>
47 #include <linux/audit.h>
48 
49 #define CREATE_TRACE_POINTS
50 #include <trace/events/signal.h>
51 
52 #include <asm/param.h>
53 #include <linux/uaccess.h>
54 #include <asm/unistd.h>
55 #include <asm/siginfo.h>
56 #include <asm/cacheflush.h>
57 
58 /*
59  * SLAB caches for signal bits.
60  */
61 
62 static struct kmem_cache *sigqueue_cachep;
63 
64 int print_fatal_signals __read_mostly;
65 
66 static void __user *sig_handler(struct task_struct *t, int sig)
67 {
68 	return t->sighand->action[sig - 1].sa.sa_handler;
69 }
70 
71 static inline bool sig_handler_ignored(void __user *handler, int sig)
72 {
73 	/* Is it explicitly or implicitly ignored? */
74 	return handler == SIG_IGN ||
75 	       (handler == SIG_DFL && sig_kernel_ignore(sig));
76 }
77 
78 static bool sig_task_ignored(struct task_struct *t, int sig, bool force)
79 {
80 	void __user *handler;
81 
82 	handler = sig_handler(t, sig);
83 
84 	/* SIGKILL and SIGSTOP may not be sent to the global init */
85 	if (unlikely(is_global_init(t) && sig_kernel_only(sig)))
86 		return true;
87 
88 	if (unlikely(t->signal->flags & SIGNAL_UNKILLABLE) &&
89 	    handler == SIG_DFL && !(force && sig_kernel_only(sig)))
90 		return true;
91 
92 	/* Only allow kernel generated signals to this kthread */
93 	if (unlikely((t->flags & PF_KTHREAD) &&
94 		     (handler == SIG_KTHREAD_KERNEL) && !force))
95 		return true;
96 
97 	return sig_handler_ignored(handler, sig);
98 }
99 
100 static bool sig_ignored(struct task_struct *t, int sig, bool force)
101 {
102 	/*
103 	 * Blocked signals are never ignored, since the
104 	 * signal handler may change by the time it is
105 	 * unblocked.
106 	 */
107 	if (sigismember(&t->blocked, sig) || sigismember(&t->real_blocked, sig))
108 		return false;
109 
110 	/*
111 	 * Tracers may want to know about even ignored signal unless it
112 	 * is SIGKILL which can't be reported anyway but can be ignored
113 	 * by SIGNAL_UNKILLABLE task.
114 	 */
115 	if (t->ptrace && sig != SIGKILL)
116 		return false;
117 
118 	return sig_task_ignored(t, sig, force);
119 }
120 
121 /*
122  * Re-calculate pending state from the set of locally pending
123  * signals, globally pending signals, and blocked signals.
124  */
125 static inline bool has_pending_signals(sigset_t *signal, sigset_t *blocked)
126 {
127 	unsigned long ready;
128 	long i;
129 
130 	switch (_NSIG_WORDS) {
131 	default:
132 		for (i = _NSIG_WORDS, ready = 0; --i >= 0 ;)
133 			ready |= signal->sig[i] &~ blocked->sig[i];
134 		break;
135 
136 	case 4: ready  = signal->sig[3] &~ blocked->sig[3];
137 		ready |= signal->sig[2] &~ blocked->sig[2];
138 		ready |= signal->sig[1] &~ blocked->sig[1];
139 		ready |= signal->sig[0] &~ blocked->sig[0];
140 		break;
141 
142 	case 2: ready  = signal->sig[1] &~ blocked->sig[1];
143 		ready |= signal->sig[0] &~ blocked->sig[0];
144 		break;
145 
146 	case 1: ready  = signal->sig[0] &~ blocked->sig[0];
147 	}
148 	return ready !=	0;
149 }
150 
151 #define PENDING(p,b) has_pending_signals(&(p)->signal, (b))
152 
153 static bool recalc_sigpending_tsk(struct task_struct *t)
154 {
155 	if ((t->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) ||
156 	    PENDING(&t->pending, &t->blocked) ||
157 	    PENDING(&t->signal->shared_pending, &t->blocked) ||
158 	    cgroup_task_frozen(t)) {
159 		set_tsk_thread_flag(t, TIF_SIGPENDING);
160 		return true;
161 	}
162 
163 	/*
164 	 * We must never clear the flag in another thread, or in current
165 	 * when it's possible the current syscall is returning -ERESTART*.
166 	 * So we don't clear it here, and only callers who know they should do.
167 	 */
168 	return false;
169 }
170 
171 /*
172  * After recalculating TIF_SIGPENDING, we need to make sure the task wakes up.
173  * This is superfluous when called on current, the wakeup is a harmless no-op.
174  */
175 void recalc_sigpending_and_wake(struct task_struct *t)
176 {
177 	if (recalc_sigpending_tsk(t))
178 		signal_wake_up(t, 0);
179 }
180 
181 void recalc_sigpending(void)
182 {
183 	if (!recalc_sigpending_tsk(current) && !freezing(current))
184 		clear_thread_flag(TIF_SIGPENDING);
185 
186 }
187 EXPORT_SYMBOL(recalc_sigpending);
188 
189 void calculate_sigpending(void)
190 {
191 	/* Have any signals or users of TIF_SIGPENDING been delayed
192 	 * until after fork?
193 	 */
194 	spin_lock_irq(&current->sighand->siglock);
195 	set_tsk_thread_flag(current, TIF_SIGPENDING);
196 	recalc_sigpending();
197 	spin_unlock_irq(&current->sighand->siglock);
198 }
199 
200 /* Given the mask, find the first available signal that should be serviced. */
201 
202 #define SYNCHRONOUS_MASK \
203 	(sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
204 	 sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
205 
206 int next_signal(struct sigpending *pending, sigset_t *mask)
207 {
208 	unsigned long i, *s, *m, x;
209 	int sig = 0;
210 
211 	s = pending->signal.sig;
212 	m = mask->sig;
213 
214 	/*
215 	 * Handle the first word specially: it contains the
216 	 * synchronous signals that need to be dequeued first.
217 	 */
218 	x = *s &~ *m;
219 	if (x) {
220 		if (x & SYNCHRONOUS_MASK)
221 			x &= SYNCHRONOUS_MASK;
222 		sig = ffz(~x) + 1;
223 		return sig;
224 	}
225 
226 	switch (_NSIG_WORDS) {
227 	default:
228 		for (i = 1; i < _NSIG_WORDS; ++i) {
229 			x = *++s &~ *++m;
230 			if (!x)
231 				continue;
232 			sig = ffz(~x) + i*_NSIG_BPW + 1;
233 			break;
234 		}
235 		break;
236 
237 	case 2:
238 		x = s[1] &~ m[1];
239 		if (!x)
240 			break;
241 		sig = ffz(~x) + _NSIG_BPW + 1;
242 		break;
243 
244 	case 1:
245 		/* Nothing to do */
246 		break;
247 	}
248 
249 	return sig;
250 }
251 
252 static inline void print_dropped_signal(int sig)
253 {
254 	static DEFINE_RATELIMIT_STATE(ratelimit_state, 5 * HZ, 10);
255 
256 	if (!print_fatal_signals)
257 		return;
258 
259 	if (!__ratelimit(&ratelimit_state))
260 		return;
261 
262 	pr_info("%s/%d: reached RLIMIT_SIGPENDING, dropped signal %d\n",
263 				current->comm, current->pid, sig);
264 }
265 
266 /**
267  * task_set_jobctl_pending - set jobctl pending bits
268  * @task: target task
269  * @mask: pending bits to set
270  *
271  * Clear @mask from @task->jobctl.  @mask must be subset of
272  * %JOBCTL_PENDING_MASK | %JOBCTL_STOP_CONSUME | %JOBCTL_STOP_SIGMASK |
273  * %JOBCTL_TRAPPING.  If stop signo is being set, the existing signo is
274  * cleared.  If @task is already being killed or exiting, this function
275  * becomes noop.
276  *
277  * CONTEXT:
278  * Must be called with @task->sighand->siglock held.
279  *
280  * RETURNS:
281  * %true if @mask is set, %false if made noop because @task was dying.
282  */
283 bool task_set_jobctl_pending(struct task_struct *task, unsigned long mask)
284 {
285 	BUG_ON(mask & ~(JOBCTL_PENDING_MASK | JOBCTL_STOP_CONSUME |
286 			JOBCTL_STOP_SIGMASK | JOBCTL_TRAPPING));
287 	BUG_ON((mask & JOBCTL_TRAPPING) && !(mask & JOBCTL_PENDING_MASK));
288 
289 	if (unlikely(fatal_signal_pending(task) || (task->flags & PF_EXITING)))
290 		return false;
291 
292 	if (mask & JOBCTL_STOP_SIGMASK)
293 		task->jobctl &= ~JOBCTL_STOP_SIGMASK;
294 
295 	task->jobctl |= mask;
296 	return true;
297 }
298 
299 /**
300  * task_clear_jobctl_trapping - clear jobctl trapping bit
301  * @task: target task
302  *
303  * If JOBCTL_TRAPPING is set, a ptracer is waiting for us to enter TRACED.
304  * Clear it and wake up the ptracer.  Note that we don't need any further
305  * locking.  @task->siglock guarantees that @task->parent points to the
306  * ptracer.
307  *
308  * CONTEXT:
309  * Must be called with @task->sighand->siglock held.
310  */
311 void task_clear_jobctl_trapping(struct task_struct *task)
312 {
313 	if (unlikely(task->jobctl & JOBCTL_TRAPPING)) {
314 		task->jobctl &= ~JOBCTL_TRAPPING;
315 		smp_mb();	/* advised by wake_up_bit() */
316 		wake_up_bit(&task->jobctl, JOBCTL_TRAPPING_BIT);
317 	}
318 }
319 
320 /**
321  * task_clear_jobctl_pending - clear jobctl pending bits
322  * @task: target task
323  * @mask: pending bits to clear
324  *
325  * Clear @mask from @task->jobctl.  @mask must be subset of
326  * %JOBCTL_PENDING_MASK.  If %JOBCTL_STOP_PENDING is being cleared, other
327  * STOP bits are cleared together.
328  *
329  * If clearing of @mask leaves no stop or trap pending, this function calls
330  * task_clear_jobctl_trapping().
331  *
332  * CONTEXT:
333  * Must be called with @task->sighand->siglock held.
334  */
335 void task_clear_jobctl_pending(struct task_struct *task, unsigned long mask)
336 {
337 	BUG_ON(mask & ~JOBCTL_PENDING_MASK);
338 
339 	if (mask & JOBCTL_STOP_PENDING)
340 		mask |= JOBCTL_STOP_CONSUME | JOBCTL_STOP_DEQUEUED;
341 
342 	task->jobctl &= ~mask;
343 
344 	if (!(task->jobctl & JOBCTL_PENDING_MASK))
345 		task_clear_jobctl_trapping(task);
346 }
347 
348 /**
349  * task_participate_group_stop - participate in a group stop
350  * @task: task participating in a group stop
351  *
352  * @task has %JOBCTL_STOP_PENDING set and is participating in a group stop.
353  * Group stop states are cleared and the group stop count is consumed if
354  * %JOBCTL_STOP_CONSUME was set.  If the consumption completes the group
355  * stop, the appropriate `SIGNAL_*` flags are set.
356  *
357  * CONTEXT:
358  * Must be called with @task->sighand->siglock held.
359  *
360  * RETURNS:
361  * %true if group stop completion should be notified to the parent, %false
362  * otherwise.
363  */
364 static bool task_participate_group_stop(struct task_struct *task)
365 {
366 	struct signal_struct *sig = task->signal;
367 	bool consume = task->jobctl & JOBCTL_STOP_CONSUME;
368 
369 	WARN_ON_ONCE(!(task->jobctl & JOBCTL_STOP_PENDING));
370 
371 	task_clear_jobctl_pending(task, JOBCTL_STOP_PENDING);
372 
373 	if (!consume)
374 		return false;
375 
376 	if (!WARN_ON_ONCE(sig->group_stop_count == 0))
377 		sig->group_stop_count--;
378 
379 	/*
380 	 * Tell the caller to notify completion iff we are entering into a
381 	 * fresh group stop.  Read comment in do_signal_stop() for details.
382 	 */
383 	if (!sig->group_stop_count && !(sig->flags & SIGNAL_STOP_STOPPED)) {
384 		signal_set_stop_flags(sig, SIGNAL_STOP_STOPPED);
385 		return true;
386 	}
387 	return false;
388 }
389 
390 void task_join_group_stop(struct task_struct *task)
391 {
392 	unsigned long mask = current->jobctl & JOBCTL_STOP_SIGMASK;
393 	struct signal_struct *sig = current->signal;
394 
395 	if (sig->group_stop_count) {
396 		sig->group_stop_count++;
397 		mask |= JOBCTL_STOP_CONSUME;
398 	} else if (!(sig->flags & SIGNAL_STOP_STOPPED))
399 		return;
400 
401 	/* Have the new thread join an on-going signal group stop */
402 	task_set_jobctl_pending(task, mask | JOBCTL_STOP_PENDING);
403 }
404 
405 /*
406  * allocate a new signal queue record
407  * - this may be called without locks if and only if t == current, otherwise an
408  *   appropriate lock must be held to stop the target task from exiting
409  */
410 static struct sigqueue *
411 __sigqueue_alloc(int sig, struct task_struct *t, gfp_t flags, int override_rlimit)
412 {
413 	struct sigqueue *q = NULL;
414 	struct user_struct *user;
415 	int sigpending;
416 
417 	/*
418 	 * Protect access to @t credentials. This can go away when all
419 	 * callers hold rcu read lock.
420 	 *
421 	 * NOTE! A pending signal will hold on to the user refcount,
422 	 * and we get/put the refcount only when the sigpending count
423 	 * changes from/to zero.
424 	 */
425 	rcu_read_lock();
426 	user = __task_cred(t)->user;
427 	sigpending = atomic_inc_return(&user->sigpending);
428 	if (sigpending == 1)
429 		get_uid(user);
430 	rcu_read_unlock();
431 
432 	if (override_rlimit || likely(sigpending <= task_rlimit(t, RLIMIT_SIGPENDING))) {
433 		q = kmem_cache_alloc(sigqueue_cachep, flags);
434 	} else {
435 		print_dropped_signal(sig);
436 	}
437 
438 	if (unlikely(q == NULL)) {
439 		if (atomic_dec_and_test(&user->sigpending))
440 			free_uid(user);
441 	} else {
442 		INIT_LIST_HEAD(&q->list);
443 		q->flags = 0;
444 		q->user = user;
445 	}
446 
447 	return q;
448 }
449 
450 static void __sigqueue_free(struct sigqueue *q)
451 {
452 	if (q->flags & SIGQUEUE_PREALLOC)
453 		return;
454 	if (atomic_dec_and_test(&q->user->sigpending))
455 		free_uid(q->user);
456 	kmem_cache_free(sigqueue_cachep, q);
457 }
458 
459 void flush_sigqueue(struct sigpending *queue)
460 {
461 	struct sigqueue *q;
462 
463 	sigemptyset(&queue->signal);
464 	while (!list_empty(&queue->list)) {
465 		q = list_entry(queue->list.next, struct sigqueue , list);
466 		list_del_init(&q->list);
467 		__sigqueue_free(q);
468 	}
469 }
470 
471 /*
472  * Flush all pending signals for this kthread.
473  */
474 void flush_signals(struct task_struct *t)
475 {
476 	unsigned long flags;
477 
478 	spin_lock_irqsave(&t->sighand->siglock, flags);
479 	clear_tsk_thread_flag(t, TIF_SIGPENDING);
480 	flush_sigqueue(&t->pending);
481 	flush_sigqueue(&t->signal->shared_pending);
482 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
483 }
484 EXPORT_SYMBOL(flush_signals);
485 
486 #ifdef CONFIG_POSIX_TIMERS
487 static void __flush_itimer_signals(struct sigpending *pending)
488 {
489 	sigset_t signal, retain;
490 	struct sigqueue *q, *n;
491 
492 	signal = pending->signal;
493 	sigemptyset(&retain);
494 
495 	list_for_each_entry_safe(q, n, &pending->list, list) {
496 		int sig = q->info.si_signo;
497 
498 		if (likely(q->info.si_code != SI_TIMER)) {
499 			sigaddset(&retain, sig);
500 		} else {
501 			sigdelset(&signal, sig);
502 			list_del_init(&q->list);
503 			__sigqueue_free(q);
504 		}
505 	}
506 
507 	sigorsets(&pending->signal, &signal, &retain);
508 }
509 
510 void flush_itimer_signals(void)
511 {
512 	struct task_struct *tsk = current;
513 	unsigned long flags;
514 
515 	spin_lock_irqsave(&tsk->sighand->siglock, flags);
516 	__flush_itimer_signals(&tsk->pending);
517 	__flush_itimer_signals(&tsk->signal->shared_pending);
518 	spin_unlock_irqrestore(&tsk->sighand->siglock, flags);
519 }
520 #endif
521 
522 void ignore_signals(struct task_struct *t)
523 {
524 	int i;
525 
526 	for (i = 0; i < _NSIG; ++i)
527 		t->sighand->action[i].sa.sa_handler = SIG_IGN;
528 
529 	flush_signals(t);
530 }
531 
532 /*
533  * Flush all handlers for a task.
534  */
535 
536 void
537 flush_signal_handlers(struct task_struct *t, int force_default)
538 {
539 	int i;
540 	struct k_sigaction *ka = &t->sighand->action[0];
541 	for (i = _NSIG ; i != 0 ; i--) {
542 		if (force_default || ka->sa.sa_handler != SIG_IGN)
543 			ka->sa.sa_handler = SIG_DFL;
544 		ka->sa.sa_flags = 0;
545 #ifdef __ARCH_HAS_SA_RESTORER
546 		ka->sa.sa_restorer = NULL;
547 #endif
548 		sigemptyset(&ka->sa.sa_mask);
549 		ka++;
550 	}
551 }
552 
553 bool unhandled_signal(struct task_struct *tsk, int sig)
554 {
555 	void __user *handler = tsk->sighand->action[sig-1].sa.sa_handler;
556 	if (is_global_init(tsk))
557 		return true;
558 
559 	if (handler != SIG_IGN && handler != SIG_DFL)
560 		return false;
561 
562 	/* if ptraced, let the tracer determine */
563 	return !tsk->ptrace;
564 }
565 
566 static void collect_signal(int sig, struct sigpending *list, kernel_siginfo_t *info,
567 			   bool *resched_timer)
568 {
569 	struct sigqueue *q, *first = NULL;
570 
571 	/*
572 	 * Collect the siginfo appropriate to this signal.  Check if
573 	 * there is another siginfo for the same signal.
574 	*/
575 	list_for_each_entry(q, &list->list, list) {
576 		if (q->info.si_signo == sig) {
577 			if (first)
578 				goto still_pending;
579 			first = q;
580 		}
581 	}
582 
583 	sigdelset(&list->signal, sig);
584 
585 	if (first) {
586 still_pending:
587 		list_del_init(&first->list);
588 		copy_siginfo(info, &first->info);
589 
590 		*resched_timer =
591 			(first->flags & SIGQUEUE_PREALLOC) &&
592 			(info->si_code == SI_TIMER) &&
593 			(info->si_sys_private);
594 
595 		__sigqueue_free(first);
596 	} else {
597 		/*
598 		 * Ok, it wasn't in the queue.  This must be
599 		 * a fast-pathed signal or we must have been
600 		 * out of queue space.  So zero out the info.
601 		 */
602 		clear_siginfo(info);
603 		info->si_signo = sig;
604 		info->si_errno = 0;
605 		info->si_code = SI_USER;
606 		info->si_pid = 0;
607 		info->si_uid = 0;
608 	}
609 }
610 
611 static int __dequeue_signal(struct sigpending *pending, sigset_t *mask,
612 			kernel_siginfo_t *info, bool *resched_timer)
613 {
614 	int sig = next_signal(pending, mask);
615 
616 	if (sig)
617 		collect_signal(sig, pending, info, resched_timer);
618 	return sig;
619 }
620 
621 /*
622  * Dequeue a signal and return the element to the caller, which is
623  * expected to free it.
624  *
625  * All callers have to hold the siglock.
626  */
627 int dequeue_signal(struct task_struct *tsk, sigset_t *mask, kernel_siginfo_t *info)
628 {
629 	bool resched_timer = false;
630 	int signr;
631 
632 	/* We only dequeue private signals from ourselves, we don't let
633 	 * signalfd steal them
634 	 */
635 	signr = __dequeue_signal(&tsk->pending, mask, info, &resched_timer);
636 	if (!signr) {
637 		signr = __dequeue_signal(&tsk->signal->shared_pending,
638 					 mask, info, &resched_timer);
639 #ifdef CONFIG_POSIX_TIMERS
640 		/*
641 		 * itimer signal ?
642 		 *
643 		 * itimers are process shared and we restart periodic
644 		 * itimers in the signal delivery path to prevent DoS
645 		 * attacks in the high resolution timer case. This is
646 		 * compliant with the old way of self-restarting
647 		 * itimers, as the SIGALRM is a legacy signal and only
648 		 * queued once. Changing the restart behaviour to
649 		 * restart the timer in the signal dequeue path is
650 		 * reducing the timer noise on heavy loaded !highres
651 		 * systems too.
652 		 */
653 		if (unlikely(signr == SIGALRM)) {
654 			struct hrtimer *tmr = &tsk->signal->real_timer;
655 
656 			if (!hrtimer_is_queued(tmr) &&
657 			    tsk->signal->it_real_incr != 0) {
658 				hrtimer_forward(tmr, tmr->base->get_time(),
659 						tsk->signal->it_real_incr);
660 				hrtimer_restart(tmr);
661 			}
662 		}
663 #endif
664 	}
665 
666 	recalc_sigpending();
667 	if (!signr)
668 		return 0;
669 
670 	if (unlikely(sig_kernel_stop(signr))) {
671 		/*
672 		 * Set a marker that we have dequeued a stop signal.  Our
673 		 * caller might release the siglock and then the pending
674 		 * stop signal it is about to process is no longer in the
675 		 * pending bitmasks, but must still be cleared by a SIGCONT
676 		 * (and overruled by a SIGKILL).  So those cases clear this
677 		 * shared flag after we've set it.  Note that this flag may
678 		 * remain set after the signal we return is ignored or
679 		 * handled.  That doesn't matter because its only purpose
680 		 * is to alert stop-signal processing code when another
681 		 * processor has come along and cleared the flag.
682 		 */
683 		current->jobctl |= JOBCTL_STOP_DEQUEUED;
684 	}
685 #ifdef CONFIG_POSIX_TIMERS
686 	if (resched_timer) {
687 		/*
688 		 * Release the siglock to ensure proper locking order
689 		 * of timer locks outside of siglocks.  Note, we leave
690 		 * irqs disabled here, since the posix-timers code is
691 		 * about to disable them again anyway.
692 		 */
693 		spin_unlock(&tsk->sighand->siglock);
694 		posixtimer_rearm(info);
695 		spin_lock(&tsk->sighand->siglock);
696 
697 		/* Don't expose the si_sys_private value to userspace */
698 		info->si_sys_private = 0;
699 	}
700 #endif
701 	return signr;
702 }
703 EXPORT_SYMBOL_GPL(dequeue_signal);
704 
705 static int dequeue_synchronous_signal(kernel_siginfo_t *info)
706 {
707 	struct task_struct *tsk = current;
708 	struct sigpending *pending = &tsk->pending;
709 	struct sigqueue *q, *sync = NULL;
710 
711 	/*
712 	 * Might a synchronous signal be in the queue?
713 	 */
714 	if (!((pending->signal.sig[0] & ~tsk->blocked.sig[0]) & SYNCHRONOUS_MASK))
715 		return 0;
716 
717 	/*
718 	 * Return the first synchronous signal in the queue.
719 	 */
720 	list_for_each_entry(q, &pending->list, list) {
721 		/* Synchronous signals have a positive si_code */
722 		if ((q->info.si_code > SI_USER) &&
723 		    (sigmask(q->info.si_signo) & SYNCHRONOUS_MASK)) {
724 			sync = q;
725 			goto next;
726 		}
727 	}
728 	return 0;
729 next:
730 	/*
731 	 * Check if there is another siginfo for the same signal.
732 	 */
733 	list_for_each_entry_continue(q, &pending->list, list) {
734 		if (q->info.si_signo == sync->info.si_signo)
735 			goto still_pending;
736 	}
737 
738 	sigdelset(&pending->signal, sync->info.si_signo);
739 	recalc_sigpending();
740 still_pending:
741 	list_del_init(&sync->list);
742 	copy_siginfo(info, &sync->info);
743 	__sigqueue_free(sync);
744 	return info->si_signo;
745 }
746 
747 /*
748  * Tell a process that it has a new active signal..
749  *
750  * NOTE! we rely on the previous spin_lock to
751  * lock interrupts for us! We can only be called with
752  * "siglock" held, and the local interrupt must
753  * have been disabled when that got acquired!
754  *
755  * No need to set need_resched since signal event passing
756  * goes through ->blocked
757  */
758 void signal_wake_up_state(struct task_struct *t, unsigned int state)
759 {
760 	set_tsk_thread_flag(t, TIF_SIGPENDING);
761 	/*
762 	 * TASK_WAKEKILL also means wake it up in the stopped/traced/killable
763 	 * case. We don't check t->state here because there is a race with it
764 	 * executing another processor and just now entering stopped state.
765 	 * By using wake_up_state, we ensure the process will wake up and
766 	 * handle its death signal.
767 	 */
768 	if (!wake_up_state(t, state | TASK_INTERRUPTIBLE))
769 		kick_process(t);
770 }
771 
772 /*
773  * Remove signals in mask from the pending set and queue.
774  * Returns 1 if any signals were found.
775  *
776  * All callers must be holding the siglock.
777  */
778 static void flush_sigqueue_mask(sigset_t *mask, struct sigpending *s)
779 {
780 	struct sigqueue *q, *n;
781 	sigset_t m;
782 
783 	sigandsets(&m, mask, &s->signal);
784 	if (sigisemptyset(&m))
785 		return;
786 
787 	sigandnsets(&s->signal, &s->signal, mask);
788 	list_for_each_entry_safe(q, n, &s->list, list) {
789 		if (sigismember(mask, q->info.si_signo)) {
790 			list_del_init(&q->list);
791 			__sigqueue_free(q);
792 		}
793 	}
794 }
795 
796 static inline int is_si_special(const struct kernel_siginfo *info)
797 {
798 	return info <= SEND_SIG_PRIV;
799 }
800 
801 static inline bool si_fromuser(const struct kernel_siginfo *info)
802 {
803 	return info == SEND_SIG_NOINFO ||
804 		(!is_si_special(info) && SI_FROMUSER(info));
805 }
806 
807 /*
808  * called with RCU read lock from check_kill_permission()
809  */
810 static bool kill_ok_by_cred(struct task_struct *t)
811 {
812 	const struct cred *cred = current_cred();
813 	const struct cred *tcred = __task_cred(t);
814 
815 	return uid_eq(cred->euid, tcred->suid) ||
816 	       uid_eq(cred->euid, tcred->uid) ||
817 	       uid_eq(cred->uid, tcred->suid) ||
818 	       uid_eq(cred->uid, tcred->uid) ||
819 	       ns_capable(tcred->user_ns, CAP_KILL);
820 }
821 
822 /*
823  * Bad permissions for sending the signal
824  * - the caller must hold the RCU read lock
825  */
826 static int check_kill_permission(int sig, struct kernel_siginfo *info,
827 				 struct task_struct *t)
828 {
829 	struct pid *sid;
830 	int error;
831 
832 	if (!valid_signal(sig))
833 		return -EINVAL;
834 
835 	if (!si_fromuser(info))
836 		return 0;
837 
838 	error = audit_signal_info(sig, t); /* Let audit system see the signal */
839 	if (error)
840 		return error;
841 
842 	if (!same_thread_group(current, t) &&
843 	    !kill_ok_by_cred(t)) {
844 		switch (sig) {
845 		case SIGCONT:
846 			sid = task_session(t);
847 			/*
848 			 * We don't return the error if sid == NULL. The
849 			 * task was unhashed, the caller must notice this.
850 			 */
851 			if (!sid || sid == task_session(current))
852 				break;
853 			fallthrough;
854 		default:
855 			return -EPERM;
856 		}
857 	}
858 
859 	return security_task_kill(t, info, sig, NULL);
860 }
861 
862 /**
863  * ptrace_trap_notify - schedule trap to notify ptracer
864  * @t: tracee wanting to notify tracer
865  *
866  * This function schedules sticky ptrace trap which is cleared on the next
867  * TRAP_STOP to notify ptracer of an event.  @t must have been seized by
868  * ptracer.
869  *
870  * If @t is running, STOP trap will be taken.  If trapped for STOP and
871  * ptracer is listening for events, tracee is woken up so that it can
872  * re-trap for the new event.  If trapped otherwise, STOP trap will be
873  * eventually taken without returning to userland after the existing traps
874  * are finished by PTRACE_CONT.
875  *
876  * CONTEXT:
877  * Must be called with @task->sighand->siglock held.
878  */
879 static void ptrace_trap_notify(struct task_struct *t)
880 {
881 	WARN_ON_ONCE(!(t->ptrace & PT_SEIZED));
882 	assert_spin_locked(&t->sighand->siglock);
883 
884 	task_set_jobctl_pending(t, JOBCTL_TRAP_NOTIFY);
885 	ptrace_signal_wake_up(t, t->jobctl & JOBCTL_LISTENING);
886 }
887 
888 /*
889  * Handle magic process-wide effects of stop/continue signals. Unlike
890  * the signal actions, these happen immediately at signal-generation
891  * time regardless of blocking, ignoring, or handling.  This does the
892  * actual continuing for SIGCONT, but not the actual stopping for stop
893  * signals. The process stop is done as a signal action for SIG_DFL.
894  *
895  * Returns true if the signal should be actually delivered, otherwise
896  * it should be dropped.
897  */
898 static bool prepare_signal(int sig, struct task_struct *p, bool force)
899 {
900 	struct signal_struct *signal = p->signal;
901 	struct task_struct *t;
902 	sigset_t flush;
903 
904 	if (signal->flags & (SIGNAL_GROUP_EXIT | SIGNAL_GROUP_COREDUMP)) {
905 		if (!(signal->flags & SIGNAL_GROUP_EXIT))
906 			return sig == SIGKILL;
907 		/*
908 		 * The process is in the middle of dying, nothing to do.
909 		 */
910 	} else if (sig_kernel_stop(sig)) {
911 		/*
912 		 * This is a stop signal.  Remove SIGCONT from all queues.
913 		 */
914 		siginitset(&flush, sigmask(SIGCONT));
915 		flush_sigqueue_mask(&flush, &signal->shared_pending);
916 		for_each_thread(p, t)
917 			flush_sigqueue_mask(&flush, &t->pending);
918 	} else if (sig == SIGCONT) {
919 		unsigned int why;
920 		/*
921 		 * Remove all stop signals from all queues, wake all threads.
922 		 */
923 		siginitset(&flush, SIG_KERNEL_STOP_MASK);
924 		flush_sigqueue_mask(&flush, &signal->shared_pending);
925 		for_each_thread(p, t) {
926 			flush_sigqueue_mask(&flush, &t->pending);
927 			task_clear_jobctl_pending(t, JOBCTL_STOP_PENDING);
928 			if (likely(!(t->ptrace & PT_SEIZED)))
929 				wake_up_state(t, __TASK_STOPPED);
930 			else
931 				ptrace_trap_notify(t);
932 		}
933 
934 		/*
935 		 * Notify the parent with CLD_CONTINUED if we were stopped.
936 		 *
937 		 * If we were in the middle of a group stop, we pretend it
938 		 * was already finished, and then continued. Since SIGCHLD
939 		 * doesn't queue we report only CLD_STOPPED, as if the next
940 		 * CLD_CONTINUED was dropped.
941 		 */
942 		why = 0;
943 		if (signal->flags & SIGNAL_STOP_STOPPED)
944 			why |= SIGNAL_CLD_CONTINUED;
945 		else if (signal->group_stop_count)
946 			why |= SIGNAL_CLD_STOPPED;
947 
948 		if (why) {
949 			/*
950 			 * The first thread which returns from do_signal_stop()
951 			 * will take ->siglock, notice SIGNAL_CLD_MASK, and
952 			 * notify its parent. See get_signal().
953 			 */
954 			signal_set_stop_flags(signal, why | SIGNAL_STOP_CONTINUED);
955 			signal->group_stop_count = 0;
956 			signal->group_exit_code = 0;
957 		}
958 	}
959 
960 	return !sig_ignored(p, sig, force);
961 }
962 
963 /*
964  * Test if P wants to take SIG.  After we've checked all threads with this,
965  * it's equivalent to finding no threads not blocking SIG.  Any threads not
966  * blocking SIG were ruled out because they are not running and already
967  * have pending signals.  Such threads will dequeue from the shared queue
968  * as soon as they're available, so putting the signal on the shared queue
969  * will be equivalent to sending it to one such thread.
970  */
971 static inline bool wants_signal(int sig, struct task_struct *p)
972 {
973 	if (sigismember(&p->blocked, sig))
974 		return false;
975 
976 	if (p->flags & PF_EXITING)
977 		return false;
978 
979 	if (sig == SIGKILL)
980 		return true;
981 
982 	if (task_is_stopped_or_traced(p))
983 		return false;
984 
985 	return task_curr(p) || !task_sigpending(p);
986 }
987 
988 static void complete_signal(int sig, struct task_struct *p, enum pid_type type)
989 {
990 	struct signal_struct *signal = p->signal;
991 	struct task_struct *t;
992 
993 	/*
994 	 * Now find a thread we can wake up to take the signal off the queue.
995 	 *
996 	 * If the main thread wants the signal, it gets first crack.
997 	 * Probably the least surprising to the average bear.
998 	 */
999 	if (wants_signal(sig, p))
1000 		t = p;
1001 	else if ((type == PIDTYPE_PID) || thread_group_empty(p))
1002 		/*
1003 		 * There is just one thread and it does not need to be woken.
1004 		 * It will dequeue unblocked signals before it runs again.
1005 		 */
1006 		return;
1007 	else {
1008 		/*
1009 		 * Otherwise try to find a suitable thread.
1010 		 */
1011 		t = signal->curr_target;
1012 		while (!wants_signal(sig, t)) {
1013 			t = next_thread(t);
1014 			if (t == signal->curr_target)
1015 				/*
1016 				 * No thread needs to be woken.
1017 				 * Any eligible threads will see
1018 				 * the signal in the queue soon.
1019 				 */
1020 				return;
1021 		}
1022 		signal->curr_target = t;
1023 	}
1024 
1025 	/*
1026 	 * Found a killable thread.  If the signal will be fatal,
1027 	 * then start taking the whole group down immediately.
1028 	 */
1029 	if (sig_fatal(p, sig) &&
1030 	    !(signal->flags & SIGNAL_GROUP_EXIT) &&
1031 	    !sigismember(&t->real_blocked, sig) &&
1032 	    (sig == SIGKILL || !p->ptrace)) {
1033 		/*
1034 		 * This signal will be fatal to the whole group.
1035 		 */
1036 		if (!sig_kernel_coredump(sig)) {
1037 			/*
1038 			 * Start a group exit and wake everybody up.
1039 			 * This way we don't have other threads
1040 			 * running and doing things after a slower
1041 			 * thread has the fatal signal pending.
1042 			 */
1043 			signal->flags = SIGNAL_GROUP_EXIT;
1044 			signal->group_exit_code = sig;
1045 			signal->group_stop_count = 0;
1046 			t = p;
1047 			do {
1048 				task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1049 				sigaddset(&t->pending.signal, SIGKILL);
1050 				signal_wake_up(t, 1);
1051 			} while_each_thread(p, t);
1052 			return;
1053 		}
1054 	}
1055 
1056 	/*
1057 	 * The signal is already in the shared-pending queue.
1058 	 * Tell the chosen thread to wake up and dequeue it.
1059 	 */
1060 	signal_wake_up(t, sig == SIGKILL);
1061 	return;
1062 }
1063 
1064 static inline bool legacy_queue(struct sigpending *signals, int sig)
1065 {
1066 	return (sig < SIGRTMIN) && sigismember(&signals->signal, sig);
1067 }
1068 
1069 static int __send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1070 			enum pid_type type, bool force)
1071 {
1072 	struct sigpending *pending;
1073 	struct sigqueue *q;
1074 	int override_rlimit;
1075 	int ret = 0, result;
1076 
1077 	assert_spin_locked(&t->sighand->siglock);
1078 
1079 	result = TRACE_SIGNAL_IGNORED;
1080 	if (!prepare_signal(sig, t, force))
1081 		goto ret;
1082 
1083 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1084 	/*
1085 	 * Short-circuit ignored signals and support queuing
1086 	 * exactly one non-rt signal, so that we can get more
1087 	 * detailed information about the cause of the signal.
1088 	 */
1089 	result = TRACE_SIGNAL_ALREADY_PENDING;
1090 	if (legacy_queue(pending, sig))
1091 		goto ret;
1092 
1093 	result = TRACE_SIGNAL_DELIVERED;
1094 	/*
1095 	 * Skip useless siginfo allocation for SIGKILL and kernel threads.
1096 	 */
1097 	if ((sig == SIGKILL) || (t->flags & PF_KTHREAD))
1098 		goto out_set;
1099 
1100 	/*
1101 	 * Real-time signals must be queued if sent by sigqueue, or
1102 	 * some other real-time mechanism.  It is implementation
1103 	 * defined whether kill() does so.  We attempt to do so, on
1104 	 * the principle of least surprise, but since kill is not
1105 	 * allowed to fail with EAGAIN when low on memory we just
1106 	 * make sure at least one signal gets delivered and don't
1107 	 * pass on the info struct.
1108 	 */
1109 	if (sig < SIGRTMIN)
1110 		override_rlimit = (is_si_special(info) || info->si_code >= 0);
1111 	else
1112 		override_rlimit = 0;
1113 
1114 	q = __sigqueue_alloc(sig, t, GFP_ATOMIC, override_rlimit);
1115 	if (q) {
1116 		list_add_tail(&q->list, &pending->list);
1117 		switch ((unsigned long) info) {
1118 		case (unsigned long) SEND_SIG_NOINFO:
1119 			clear_siginfo(&q->info);
1120 			q->info.si_signo = sig;
1121 			q->info.si_errno = 0;
1122 			q->info.si_code = SI_USER;
1123 			q->info.si_pid = task_tgid_nr_ns(current,
1124 							task_active_pid_ns(t));
1125 			rcu_read_lock();
1126 			q->info.si_uid =
1127 				from_kuid_munged(task_cred_xxx(t, user_ns),
1128 						 current_uid());
1129 			rcu_read_unlock();
1130 			break;
1131 		case (unsigned long) SEND_SIG_PRIV:
1132 			clear_siginfo(&q->info);
1133 			q->info.si_signo = sig;
1134 			q->info.si_errno = 0;
1135 			q->info.si_code = SI_KERNEL;
1136 			q->info.si_pid = 0;
1137 			q->info.si_uid = 0;
1138 			break;
1139 		default:
1140 			copy_siginfo(&q->info, info);
1141 			break;
1142 		}
1143 	} else if (!is_si_special(info) &&
1144 		   sig >= SIGRTMIN && info->si_code != SI_USER) {
1145 		/*
1146 		 * Queue overflow, abort.  We may abort if the
1147 		 * signal was rt and sent by user using something
1148 		 * other than kill().
1149 		 */
1150 		result = TRACE_SIGNAL_OVERFLOW_FAIL;
1151 		ret = -EAGAIN;
1152 		goto ret;
1153 	} else {
1154 		/*
1155 		 * This is a silent loss of information.  We still
1156 		 * send the signal, but the *info bits are lost.
1157 		 */
1158 		result = TRACE_SIGNAL_LOSE_INFO;
1159 	}
1160 
1161 out_set:
1162 	signalfd_notify(t, sig);
1163 	sigaddset(&pending->signal, sig);
1164 
1165 	/* Let multiprocess signals appear after on-going forks */
1166 	if (type > PIDTYPE_TGID) {
1167 		struct multiprocess_signals *delayed;
1168 		hlist_for_each_entry(delayed, &t->signal->multiprocess, node) {
1169 			sigset_t *signal = &delayed->signal;
1170 			/* Can't queue both a stop and a continue signal */
1171 			if (sig == SIGCONT)
1172 				sigdelsetmask(signal, SIG_KERNEL_STOP_MASK);
1173 			else if (sig_kernel_stop(sig))
1174 				sigdelset(signal, SIGCONT);
1175 			sigaddset(signal, sig);
1176 		}
1177 	}
1178 
1179 	complete_signal(sig, t, type);
1180 ret:
1181 	trace_signal_generate(sig, info, t, type != PIDTYPE_PID, result);
1182 	return ret;
1183 }
1184 
1185 static inline bool has_si_pid_and_uid(struct kernel_siginfo *info)
1186 {
1187 	bool ret = false;
1188 	switch (siginfo_layout(info->si_signo, info->si_code)) {
1189 	case SIL_KILL:
1190 	case SIL_CHLD:
1191 	case SIL_RT:
1192 		ret = true;
1193 		break;
1194 	case SIL_TIMER:
1195 	case SIL_POLL:
1196 	case SIL_FAULT:
1197 	case SIL_FAULT_MCEERR:
1198 	case SIL_FAULT_BNDERR:
1199 	case SIL_FAULT_PKUERR:
1200 	case SIL_SYS:
1201 		ret = false;
1202 		break;
1203 	}
1204 	return ret;
1205 }
1206 
1207 static int send_signal(int sig, struct kernel_siginfo *info, struct task_struct *t,
1208 			enum pid_type type)
1209 {
1210 	/* Should SIGKILL or SIGSTOP be received by a pid namespace init? */
1211 	bool force = false;
1212 
1213 	if (info == SEND_SIG_NOINFO) {
1214 		/* Force if sent from an ancestor pid namespace */
1215 		force = !task_pid_nr_ns(current, task_active_pid_ns(t));
1216 	} else if (info == SEND_SIG_PRIV) {
1217 		/* Don't ignore kernel generated signals */
1218 		force = true;
1219 	} else if (has_si_pid_and_uid(info)) {
1220 		/* SIGKILL and SIGSTOP is special or has ids */
1221 		struct user_namespace *t_user_ns;
1222 
1223 		rcu_read_lock();
1224 		t_user_ns = task_cred_xxx(t, user_ns);
1225 		if (current_user_ns() != t_user_ns) {
1226 			kuid_t uid = make_kuid(current_user_ns(), info->si_uid);
1227 			info->si_uid = from_kuid_munged(t_user_ns, uid);
1228 		}
1229 		rcu_read_unlock();
1230 
1231 		/* A kernel generated signal? */
1232 		force = (info->si_code == SI_KERNEL);
1233 
1234 		/* From an ancestor pid namespace? */
1235 		if (!task_pid_nr_ns(current, task_active_pid_ns(t))) {
1236 			info->si_pid = 0;
1237 			force = true;
1238 		}
1239 	}
1240 	return __send_signal(sig, info, t, type, force);
1241 }
1242 
1243 static void print_fatal_signal(int signr)
1244 {
1245 	struct pt_regs *regs = signal_pt_regs();
1246 	pr_info("potentially unexpected fatal signal %d.\n", signr);
1247 
1248 #if defined(__i386__) && !defined(__arch_um__)
1249 	pr_info("code at %08lx: ", regs->ip);
1250 	{
1251 		int i;
1252 		for (i = 0; i < 16; i++) {
1253 			unsigned char insn;
1254 
1255 			if (get_user(insn, (unsigned char *)(regs->ip + i)))
1256 				break;
1257 			pr_cont("%02x ", insn);
1258 		}
1259 	}
1260 	pr_cont("\n");
1261 #endif
1262 	preempt_disable();
1263 	show_regs(regs);
1264 	preempt_enable();
1265 }
1266 
1267 static int __init setup_print_fatal_signals(char *str)
1268 {
1269 	get_option (&str, &print_fatal_signals);
1270 
1271 	return 1;
1272 }
1273 
1274 __setup("print-fatal-signals=", setup_print_fatal_signals);
1275 
1276 int
1277 __group_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1278 {
1279 	return send_signal(sig, info, p, PIDTYPE_TGID);
1280 }
1281 
1282 int do_send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p,
1283 			enum pid_type type)
1284 {
1285 	unsigned long flags;
1286 	int ret = -ESRCH;
1287 
1288 	if (lock_task_sighand(p, &flags)) {
1289 		ret = send_signal(sig, info, p, type);
1290 		unlock_task_sighand(p, &flags);
1291 	}
1292 
1293 	return ret;
1294 }
1295 
1296 /*
1297  * Force a signal that the process can't ignore: if necessary
1298  * we unblock the signal and change any SIG_IGN to SIG_DFL.
1299  *
1300  * Note: If we unblock the signal, we always reset it to SIG_DFL,
1301  * since we do not want to have a signal handler that was blocked
1302  * be invoked when user space had explicitly blocked it.
1303  *
1304  * We don't want to have recursive SIGSEGV's etc, for example,
1305  * that is why we also clear SIGNAL_UNKILLABLE.
1306  */
1307 static int
1308 force_sig_info_to_task(struct kernel_siginfo *info, struct task_struct *t)
1309 {
1310 	unsigned long int flags;
1311 	int ret, blocked, ignored;
1312 	struct k_sigaction *action;
1313 	int sig = info->si_signo;
1314 
1315 	spin_lock_irqsave(&t->sighand->siglock, flags);
1316 	action = &t->sighand->action[sig-1];
1317 	ignored = action->sa.sa_handler == SIG_IGN;
1318 	blocked = sigismember(&t->blocked, sig);
1319 	if (blocked || ignored) {
1320 		action->sa.sa_handler = SIG_DFL;
1321 		if (blocked) {
1322 			sigdelset(&t->blocked, sig);
1323 			recalc_sigpending_and_wake(t);
1324 		}
1325 	}
1326 	/*
1327 	 * Don't clear SIGNAL_UNKILLABLE for traced tasks, users won't expect
1328 	 * debugging to leave init killable.
1329 	 */
1330 	if (action->sa.sa_handler == SIG_DFL && !t->ptrace)
1331 		t->signal->flags &= ~SIGNAL_UNKILLABLE;
1332 	ret = send_signal(sig, info, t, PIDTYPE_PID);
1333 	spin_unlock_irqrestore(&t->sighand->siglock, flags);
1334 
1335 	return ret;
1336 }
1337 
1338 int force_sig_info(struct kernel_siginfo *info)
1339 {
1340 	return force_sig_info_to_task(info, current);
1341 }
1342 
1343 /*
1344  * Nuke all other threads in the group.
1345  */
1346 int zap_other_threads(struct task_struct *p)
1347 {
1348 	struct task_struct *t = p;
1349 	int count = 0;
1350 
1351 	p->signal->group_stop_count = 0;
1352 
1353 	while_each_thread(p, t) {
1354 		task_clear_jobctl_pending(t, JOBCTL_PENDING_MASK);
1355 		count++;
1356 
1357 		/* Don't bother with already dead threads */
1358 		if (t->exit_state)
1359 			continue;
1360 		sigaddset(&t->pending.signal, SIGKILL);
1361 		signal_wake_up(t, 1);
1362 	}
1363 
1364 	return count;
1365 }
1366 
1367 struct sighand_struct *__lock_task_sighand(struct task_struct *tsk,
1368 					   unsigned long *flags)
1369 {
1370 	struct sighand_struct *sighand;
1371 
1372 	rcu_read_lock();
1373 	for (;;) {
1374 		sighand = rcu_dereference(tsk->sighand);
1375 		if (unlikely(sighand == NULL))
1376 			break;
1377 
1378 		/*
1379 		 * This sighand can be already freed and even reused, but
1380 		 * we rely on SLAB_TYPESAFE_BY_RCU and sighand_ctor() which
1381 		 * initializes ->siglock: this slab can't go away, it has
1382 		 * the same object type, ->siglock can't be reinitialized.
1383 		 *
1384 		 * We need to ensure that tsk->sighand is still the same
1385 		 * after we take the lock, we can race with de_thread() or
1386 		 * __exit_signal(). In the latter case the next iteration
1387 		 * must see ->sighand == NULL.
1388 		 */
1389 		spin_lock_irqsave(&sighand->siglock, *flags);
1390 		if (likely(sighand == rcu_access_pointer(tsk->sighand)))
1391 			break;
1392 		spin_unlock_irqrestore(&sighand->siglock, *flags);
1393 	}
1394 	rcu_read_unlock();
1395 
1396 	return sighand;
1397 }
1398 
1399 /*
1400  * send signal info to all the members of a group
1401  */
1402 int group_send_sig_info(int sig, struct kernel_siginfo *info,
1403 			struct task_struct *p, enum pid_type type)
1404 {
1405 	int ret;
1406 
1407 	rcu_read_lock();
1408 	ret = check_kill_permission(sig, info, p);
1409 	rcu_read_unlock();
1410 
1411 	if (!ret && sig)
1412 		ret = do_send_sig_info(sig, info, p, type);
1413 
1414 	return ret;
1415 }
1416 
1417 /*
1418  * __kill_pgrp_info() sends a signal to a process group: this is what the tty
1419  * control characters do (^C, ^Z etc)
1420  * - the caller must hold at least a readlock on tasklist_lock
1421  */
1422 int __kill_pgrp_info(int sig, struct kernel_siginfo *info, struct pid *pgrp)
1423 {
1424 	struct task_struct *p = NULL;
1425 	int retval, success;
1426 
1427 	success = 0;
1428 	retval = -ESRCH;
1429 	do_each_pid_task(pgrp, PIDTYPE_PGID, p) {
1430 		int err = group_send_sig_info(sig, info, p, PIDTYPE_PGID);
1431 		success |= !err;
1432 		retval = err;
1433 	} while_each_pid_task(pgrp, PIDTYPE_PGID, p);
1434 	return success ? 0 : retval;
1435 }
1436 
1437 int kill_pid_info(int sig, struct kernel_siginfo *info, struct pid *pid)
1438 {
1439 	int error = -ESRCH;
1440 	struct task_struct *p;
1441 
1442 	for (;;) {
1443 		rcu_read_lock();
1444 		p = pid_task(pid, PIDTYPE_PID);
1445 		if (p)
1446 			error = group_send_sig_info(sig, info, p, PIDTYPE_TGID);
1447 		rcu_read_unlock();
1448 		if (likely(!p || error != -ESRCH))
1449 			return error;
1450 
1451 		/*
1452 		 * The task was unhashed in between, try again.  If it
1453 		 * is dead, pid_task() will return NULL, if we race with
1454 		 * de_thread() it will find the new leader.
1455 		 */
1456 	}
1457 }
1458 
1459 static int kill_proc_info(int sig, struct kernel_siginfo *info, pid_t pid)
1460 {
1461 	int error;
1462 	rcu_read_lock();
1463 	error = kill_pid_info(sig, info, find_vpid(pid));
1464 	rcu_read_unlock();
1465 	return error;
1466 }
1467 
1468 static inline bool kill_as_cred_perm(const struct cred *cred,
1469 				     struct task_struct *target)
1470 {
1471 	const struct cred *pcred = __task_cred(target);
1472 
1473 	return uid_eq(cred->euid, pcred->suid) ||
1474 	       uid_eq(cred->euid, pcred->uid) ||
1475 	       uid_eq(cred->uid, pcred->suid) ||
1476 	       uid_eq(cred->uid, pcred->uid);
1477 }
1478 
1479 /*
1480  * The usb asyncio usage of siginfo is wrong.  The glibc support
1481  * for asyncio which uses SI_ASYNCIO assumes the layout is SIL_RT.
1482  * AKA after the generic fields:
1483  *	kernel_pid_t	si_pid;
1484  *	kernel_uid32_t	si_uid;
1485  *	sigval_t	si_value;
1486  *
1487  * Unfortunately when usb generates SI_ASYNCIO it assumes the layout
1488  * after the generic fields is:
1489  *	void __user 	*si_addr;
1490  *
1491  * This is a practical problem when there is a 64bit big endian kernel
1492  * and a 32bit userspace.  As the 32bit address will encoded in the low
1493  * 32bits of the pointer.  Those low 32bits will be stored at higher
1494  * address than appear in a 32 bit pointer.  So userspace will not
1495  * see the address it was expecting for it's completions.
1496  *
1497  * There is nothing in the encoding that can allow
1498  * copy_siginfo_to_user32 to detect this confusion of formats, so
1499  * handle this by requiring the caller of kill_pid_usb_asyncio to
1500  * notice when this situration takes place and to store the 32bit
1501  * pointer in sival_int, instead of sival_addr of the sigval_t addr
1502  * parameter.
1503  */
1504 int kill_pid_usb_asyncio(int sig, int errno, sigval_t addr,
1505 			 struct pid *pid, const struct cred *cred)
1506 {
1507 	struct kernel_siginfo info;
1508 	struct task_struct *p;
1509 	unsigned long flags;
1510 	int ret = -EINVAL;
1511 
1512 	if (!valid_signal(sig))
1513 		return ret;
1514 
1515 	clear_siginfo(&info);
1516 	info.si_signo = sig;
1517 	info.si_errno = errno;
1518 	info.si_code = SI_ASYNCIO;
1519 	*((sigval_t *)&info.si_pid) = addr;
1520 
1521 	rcu_read_lock();
1522 	p = pid_task(pid, PIDTYPE_PID);
1523 	if (!p) {
1524 		ret = -ESRCH;
1525 		goto out_unlock;
1526 	}
1527 	if (!kill_as_cred_perm(cred, p)) {
1528 		ret = -EPERM;
1529 		goto out_unlock;
1530 	}
1531 	ret = security_task_kill(p, &info, sig, cred);
1532 	if (ret)
1533 		goto out_unlock;
1534 
1535 	if (sig) {
1536 		if (lock_task_sighand(p, &flags)) {
1537 			ret = __send_signal(sig, &info, p, PIDTYPE_TGID, false);
1538 			unlock_task_sighand(p, &flags);
1539 		} else
1540 			ret = -ESRCH;
1541 	}
1542 out_unlock:
1543 	rcu_read_unlock();
1544 	return ret;
1545 }
1546 EXPORT_SYMBOL_GPL(kill_pid_usb_asyncio);
1547 
1548 /*
1549  * kill_something_info() interprets pid in interesting ways just like kill(2).
1550  *
1551  * POSIX specifies that kill(-1,sig) is unspecified, but what we have
1552  * is probably wrong.  Should make it like BSD or SYSV.
1553  */
1554 
1555 static int kill_something_info(int sig, struct kernel_siginfo *info, pid_t pid)
1556 {
1557 	int ret;
1558 
1559 	if (pid > 0)
1560 		return kill_proc_info(sig, info, pid);
1561 
1562 	/* -INT_MIN is undefined.  Exclude this case to avoid a UBSAN warning */
1563 	if (pid == INT_MIN)
1564 		return -ESRCH;
1565 
1566 	read_lock(&tasklist_lock);
1567 	if (pid != -1) {
1568 		ret = __kill_pgrp_info(sig, info,
1569 				pid ? find_vpid(-pid) : task_pgrp(current));
1570 	} else {
1571 		int retval = 0, count = 0;
1572 		struct task_struct * p;
1573 
1574 		for_each_process(p) {
1575 			if (task_pid_vnr(p) > 1 &&
1576 					!same_thread_group(p, current)) {
1577 				int err = group_send_sig_info(sig, info, p,
1578 							      PIDTYPE_MAX);
1579 				++count;
1580 				if (err != -EPERM)
1581 					retval = err;
1582 			}
1583 		}
1584 		ret = count ? retval : -ESRCH;
1585 	}
1586 	read_unlock(&tasklist_lock);
1587 
1588 	return ret;
1589 }
1590 
1591 /*
1592  * These are for backward compatibility with the rest of the kernel source.
1593  */
1594 
1595 int send_sig_info(int sig, struct kernel_siginfo *info, struct task_struct *p)
1596 {
1597 	/*
1598 	 * Make sure legacy kernel users don't send in bad values
1599 	 * (normal paths check this in check_kill_permission).
1600 	 */
1601 	if (!valid_signal(sig))
1602 		return -EINVAL;
1603 
1604 	return do_send_sig_info(sig, info, p, PIDTYPE_PID);
1605 }
1606 EXPORT_SYMBOL(send_sig_info);
1607 
1608 #define __si_special(priv) \
1609 	((priv) ? SEND_SIG_PRIV : SEND_SIG_NOINFO)
1610 
1611 int
1612 send_sig(int sig, struct task_struct *p, int priv)
1613 {
1614 	return send_sig_info(sig, __si_special(priv), p);
1615 }
1616 EXPORT_SYMBOL(send_sig);
1617 
1618 void force_sig(int sig)
1619 {
1620 	struct kernel_siginfo info;
1621 
1622 	clear_siginfo(&info);
1623 	info.si_signo = sig;
1624 	info.si_errno = 0;
1625 	info.si_code = SI_KERNEL;
1626 	info.si_pid = 0;
1627 	info.si_uid = 0;
1628 	force_sig_info(&info);
1629 }
1630 EXPORT_SYMBOL(force_sig);
1631 
1632 /*
1633  * When things go south during signal handling, we
1634  * will force a SIGSEGV. And if the signal that caused
1635  * the problem was already a SIGSEGV, we'll want to
1636  * make sure we don't even try to deliver the signal..
1637  */
1638 void force_sigsegv(int sig)
1639 {
1640 	struct task_struct *p = current;
1641 
1642 	if (sig == SIGSEGV) {
1643 		unsigned long flags;
1644 		spin_lock_irqsave(&p->sighand->siglock, flags);
1645 		p->sighand->action[sig - 1].sa.sa_handler = SIG_DFL;
1646 		spin_unlock_irqrestore(&p->sighand->siglock, flags);
1647 	}
1648 	force_sig(SIGSEGV);
1649 }
1650 
1651 int force_sig_fault_to_task(int sig, int code, void __user *addr
1652 	___ARCH_SI_TRAPNO(int trapno)
1653 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1654 	, struct task_struct *t)
1655 {
1656 	struct kernel_siginfo info;
1657 
1658 	clear_siginfo(&info);
1659 	info.si_signo = sig;
1660 	info.si_errno = 0;
1661 	info.si_code  = code;
1662 	info.si_addr  = addr;
1663 #ifdef __ARCH_SI_TRAPNO
1664 	info.si_trapno = trapno;
1665 #endif
1666 #ifdef __ia64__
1667 	info.si_imm = imm;
1668 	info.si_flags = flags;
1669 	info.si_isr = isr;
1670 #endif
1671 	return force_sig_info_to_task(&info, t);
1672 }
1673 
1674 int force_sig_fault(int sig, int code, void __user *addr
1675 	___ARCH_SI_TRAPNO(int trapno)
1676 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr))
1677 {
1678 	return force_sig_fault_to_task(sig, code, addr
1679 				       ___ARCH_SI_TRAPNO(trapno)
1680 				       ___ARCH_SI_IA64(imm, flags, isr), current);
1681 }
1682 
1683 int send_sig_fault(int sig, int code, void __user *addr
1684 	___ARCH_SI_TRAPNO(int trapno)
1685 	___ARCH_SI_IA64(int imm, unsigned int flags, unsigned long isr)
1686 	, struct task_struct *t)
1687 {
1688 	struct kernel_siginfo info;
1689 
1690 	clear_siginfo(&info);
1691 	info.si_signo = sig;
1692 	info.si_errno = 0;
1693 	info.si_code  = code;
1694 	info.si_addr  = addr;
1695 #ifdef __ARCH_SI_TRAPNO
1696 	info.si_trapno = trapno;
1697 #endif
1698 #ifdef __ia64__
1699 	info.si_imm = imm;
1700 	info.si_flags = flags;
1701 	info.si_isr = isr;
1702 #endif
1703 	return send_sig_info(info.si_signo, &info, t);
1704 }
1705 
1706 int force_sig_mceerr(int code, void __user *addr, short lsb)
1707 {
1708 	struct kernel_siginfo info;
1709 
1710 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1711 	clear_siginfo(&info);
1712 	info.si_signo = SIGBUS;
1713 	info.si_errno = 0;
1714 	info.si_code = code;
1715 	info.si_addr = addr;
1716 	info.si_addr_lsb = lsb;
1717 	return force_sig_info(&info);
1718 }
1719 
1720 int send_sig_mceerr(int code, void __user *addr, short lsb, struct task_struct *t)
1721 {
1722 	struct kernel_siginfo info;
1723 
1724 	WARN_ON((code != BUS_MCEERR_AO) && (code != BUS_MCEERR_AR));
1725 	clear_siginfo(&info);
1726 	info.si_signo = SIGBUS;
1727 	info.si_errno = 0;
1728 	info.si_code = code;
1729 	info.si_addr = addr;
1730 	info.si_addr_lsb = lsb;
1731 	return send_sig_info(info.si_signo, &info, t);
1732 }
1733 EXPORT_SYMBOL(send_sig_mceerr);
1734 
1735 int force_sig_bnderr(void __user *addr, void __user *lower, void __user *upper)
1736 {
1737 	struct kernel_siginfo info;
1738 
1739 	clear_siginfo(&info);
1740 	info.si_signo = SIGSEGV;
1741 	info.si_errno = 0;
1742 	info.si_code  = SEGV_BNDERR;
1743 	info.si_addr  = addr;
1744 	info.si_lower = lower;
1745 	info.si_upper = upper;
1746 	return force_sig_info(&info);
1747 }
1748 
1749 #ifdef SEGV_PKUERR
1750 int force_sig_pkuerr(void __user *addr, u32 pkey)
1751 {
1752 	struct kernel_siginfo info;
1753 
1754 	clear_siginfo(&info);
1755 	info.si_signo = SIGSEGV;
1756 	info.si_errno = 0;
1757 	info.si_code  = SEGV_PKUERR;
1758 	info.si_addr  = addr;
1759 	info.si_pkey  = pkey;
1760 	return force_sig_info(&info);
1761 }
1762 #endif
1763 
1764 /* For the crazy architectures that include trap information in
1765  * the errno field, instead of an actual errno value.
1766  */
1767 int force_sig_ptrace_errno_trap(int errno, void __user *addr)
1768 {
1769 	struct kernel_siginfo info;
1770 
1771 	clear_siginfo(&info);
1772 	info.si_signo = SIGTRAP;
1773 	info.si_errno = errno;
1774 	info.si_code  = TRAP_HWBKPT;
1775 	info.si_addr  = addr;
1776 	return force_sig_info(&info);
1777 }
1778 
1779 int kill_pgrp(struct pid *pid, int sig, int priv)
1780 {
1781 	int ret;
1782 
1783 	read_lock(&tasklist_lock);
1784 	ret = __kill_pgrp_info(sig, __si_special(priv), pid);
1785 	read_unlock(&tasklist_lock);
1786 
1787 	return ret;
1788 }
1789 EXPORT_SYMBOL(kill_pgrp);
1790 
1791 int kill_pid(struct pid *pid, int sig, int priv)
1792 {
1793 	return kill_pid_info(sig, __si_special(priv), pid);
1794 }
1795 EXPORT_SYMBOL(kill_pid);
1796 
1797 /*
1798  * These functions support sending signals using preallocated sigqueue
1799  * structures.  This is needed "because realtime applications cannot
1800  * afford to lose notifications of asynchronous events, like timer
1801  * expirations or I/O completions".  In the case of POSIX Timers
1802  * we allocate the sigqueue structure from the timer_create.  If this
1803  * allocation fails we are able to report the failure to the application
1804  * with an EAGAIN error.
1805  */
1806 struct sigqueue *sigqueue_alloc(void)
1807 {
1808 	struct sigqueue *q = __sigqueue_alloc(-1, current, GFP_KERNEL, 0);
1809 
1810 	if (q)
1811 		q->flags |= SIGQUEUE_PREALLOC;
1812 
1813 	return q;
1814 }
1815 
1816 void sigqueue_free(struct sigqueue *q)
1817 {
1818 	unsigned long flags;
1819 	spinlock_t *lock = &current->sighand->siglock;
1820 
1821 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1822 	/*
1823 	 * We must hold ->siglock while testing q->list
1824 	 * to serialize with collect_signal() or with
1825 	 * __exit_signal()->flush_sigqueue().
1826 	 */
1827 	spin_lock_irqsave(lock, flags);
1828 	q->flags &= ~SIGQUEUE_PREALLOC;
1829 	/*
1830 	 * If it is queued it will be freed when dequeued,
1831 	 * like the "regular" sigqueue.
1832 	 */
1833 	if (!list_empty(&q->list))
1834 		q = NULL;
1835 	spin_unlock_irqrestore(lock, flags);
1836 
1837 	if (q)
1838 		__sigqueue_free(q);
1839 }
1840 
1841 int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
1842 {
1843 	int sig = q->info.si_signo;
1844 	struct sigpending *pending;
1845 	struct task_struct *t;
1846 	unsigned long flags;
1847 	int ret, result;
1848 
1849 	BUG_ON(!(q->flags & SIGQUEUE_PREALLOC));
1850 
1851 	ret = -1;
1852 	rcu_read_lock();
1853 	t = pid_task(pid, type);
1854 	if (!t || !likely(lock_task_sighand(t, &flags)))
1855 		goto ret;
1856 
1857 	ret = 1; /* the signal is ignored */
1858 	result = TRACE_SIGNAL_IGNORED;
1859 	if (!prepare_signal(sig, t, false))
1860 		goto out;
1861 
1862 	ret = 0;
1863 	if (unlikely(!list_empty(&q->list))) {
1864 		/*
1865 		 * If an SI_TIMER entry is already queue just increment
1866 		 * the overrun count.
1867 		 */
1868 		BUG_ON(q->info.si_code != SI_TIMER);
1869 		q->info.si_overrun++;
1870 		result = TRACE_SIGNAL_ALREADY_PENDING;
1871 		goto out;
1872 	}
1873 	q->info.si_overrun = 0;
1874 
1875 	signalfd_notify(t, sig);
1876 	pending = (type != PIDTYPE_PID) ? &t->signal->shared_pending : &t->pending;
1877 	list_add_tail(&q->list, &pending->list);
1878 	sigaddset(&pending->signal, sig);
1879 	complete_signal(sig, t, type);
1880 	result = TRACE_SIGNAL_DELIVERED;
1881 out:
1882 	trace_signal_generate(sig, &q->info, t, type != PIDTYPE_PID, result);
1883 	unlock_task_sighand(t, &flags);
1884 ret:
1885 	rcu_read_unlock();
1886 	return ret;
1887 }
1888 
1889 static void do_notify_pidfd(struct task_struct *task)
1890 {
1891 	struct pid *pid;
1892 
1893 	WARN_ON(task->exit_state == 0);
1894 	pid = task_pid(task);
1895 	wake_up_all(&pid->wait_pidfd);
1896 }
1897 
1898 /*
1899  * Let a parent know about the death of a child.
1900  * For a stopped/continued status change, use do_notify_parent_cldstop instead.
1901  *
1902  * Returns true if our parent ignored us and so we've switched to
1903  * self-reaping.
1904  */
1905 bool do_notify_parent(struct task_struct *tsk, int sig)
1906 {
1907 	struct kernel_siginfo info;
1908 	unsigned long flags;
1909 	struct sighand_struct *psig;
1910 	bool autoreap = false;
1911 	u64 utime, stime;
1912 
1913 	BUG_ON(sig == -1);
1914 
1915  	/* do_notify_parent_cldstop should have been called instead.  */
1916  	BUG_ON(task_is_stopped_or_traced(tsk));
1917 
1918 	BUG_ON(!tsk->ptrace &&
1919 	       (tsk->group_leader != tsk || !thread_group_empty(tsk)));
1920 
1921 	/* Wake up all pidfd waiters */
1922 	do_notify_pidfd(tsk);
1923 
1924 	if (sig != SIGCHLD) {
1925 		/*
1926 		 * This is only possible if parent == real_parent.
1927 		 * Check if it has changed security domain.
1928 		 */
1929 		if (tsk->parent_exec_id != READ_ONCE(tsk->parent->self_exec_id))
1930 			sig = SIGCHLD;
1931 	}
1932 
1933 	clear_siginfo(&info);
1934 	info.si_signo = sig;
1935 	info.si_errno = 0;
1936 	/*
1937 	 * We are under tasklist_lock here so our parent is tied to
1938 	 * us and cannot change.
1939 	 *
1940 	 * task_active_pid_ns will always return the same pid namespace
1941 	 * until a task passes through release_task.
1942 	 *
1943 	 * write_lock() currently calls preempt_disable() which is the
1944 	 * same as rcu_read_lock(), but according to Oleg, this is not
1945 	 * correct to rely on this
1946 	 */
1947 	rcu_read_lock();
1948 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(tsk->parent));
1949 	info.si_uid = from_kuid_munged(task_cred_xxx(tsk->parent, user_ns),
1950 				       task_uid(tsk));
1951 	rcu_read_unlock();
1952 
1953 	task_cputime(tsk, &utime, &stime);
1954 	info.si_utime = nsec_to_clock_t(utime + tsk->signal->utime);
1955 	info.si_stime = nsec_to_clock_t(stime + tsk->signal->stime);
1956 
1957 	info.si_status = tsk->exit_code & 0x7f;
1958 	if (tsk->exit_code & 0x80)
1959 		info.si_code = CLD_DUMPED;
1960 	else if (tsk->exit_code & 0x7f)
1961 		info.si_code = CLD_KILLED;
1962 	else {
1963 		info.si_code = CLD_EXITED;
1964 		info.si_status = tsk->exit_code >> 8;
1965 	}
1966 
1967 	psig = tsk->parent->sighand;
1968 	spin_lock_irqsave(&psig->siglock, flags);
1969 	if (!tsk->ptrace && sig == SIGCHLD &&
1970 	    (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN ||
1971 	     (psig->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDWAIT))) {
1972 		/*
1973 		 * We are exiting and our parent doesn't care.  POSIX.1
1974 		 * defines special semantics for setting SIGCHLD to SIG_IGN
1975 		 * or setting the SA_NOCLDWAIT flag: we should be reaped
1976 		 * automatically and not left for our parent's wait4 call.
1977 		 * Rather than having the parent do it as a magic kind of
1978 		 * signal handler, we just set this to tell do_exit that we
1979 		 * can be cleaned up without becoming a zombie.  Note that
1980 		 * we still call __wake_up_parent in this case, because a
1981 		 * blocked sys_wait4 might now return -ECHILD.
1982 		 *
1983 		 * Whether we send SIGCHLD or not for SA_NOCLDWAIT
1984 		 * is implementation-defined: we do (if you don't want
1985 		 * it, just use SIG_IGN instead).
1986 		 */
1987 		autoreap = true;
1988 		if (psig->action[SIGCHLD-1].sa.sa_handler == SIG_IGN)
1989 			sig = 0;
1990 	}
1991 	/*
1992 	 * Send with __send_signal as si_pid and si_uid are in the
1993 	 * parent's namespaces.
1994 	 */
1995 	if (valid_signal(sig) && sig)
1996 		__send_signal(sig, &info, tsk->parent, PIDTYPE_TGID, false);
1997 	__wake_up_parent(tsk, tsk->parent);
1998 	spin_unlock_irqrestore(&psig->siglock, flags);
1999 
2000 	return autoreap;
2001 }
2002 
2003 /**
2004  * do_notify_parent_cldstop - notify parent of stopped/continued state change
2005  * @tsk: task reporting the state change
2006  * @for_ptracer: the notification is for ptracer
2007  * @why: CLD_{CONTINUED|STOPPED|TRAPPED} to report
2008  *
2009  * Notify @tsk's parent that the stopped/continued state has changed.  If
2010  * @for_ptracer is %false, @tsk's group leader notifies to its real parent.
2011  * If %true, @tsk reports to @tsk->parent which should be the ptracer.
2012  *
2013  * CONTEXT:
2014  * Must be called with tasklist_lock at least read locked.
2015  */
2016 static void do_notify_parent_cldstop(struct task_struct *tsk,
2017 				     bool for_ptracer, int why)
2018 {
2019 	struct kernel_siginfo info;
2020 	unsigned long flags;
2021 	struct task_struct *parent;
2022 	struct sighand_struct *sighand;
2023 	u64 utime, stime;
2024 
2025 	if (for_ptracer) {
2026 		parent = tsk->parent;
2027 	} else {
2028 		tsk = tsk->group_leader;
2029 		parent = tsk->real_parent;
2030 	}
2031 
2032 	clear_siginfo(&info);
2033 	info.si_signo = SIGCHLD;
2034 	info.si_errno = 0;
2035 	/*
2036 	 * see comment in do_notify_parent() about the following 4 lines
2037 	 */
2038 	rcu_read_lock();
2039 	info.si_pid = task_pid_nr_ns(tsk, task_active_pid_ns(parent));
2040 	info.si_uid = from_kuid_munged(task_cred_xxx(parent, user_ns), task_uid(tsk));
2041 	rcu_read_unlock();
2042 
2043 	task_cputime(tsk, &utime, &stime);
2044 	info.si_utime = nsec_to_clock_t(utime);
2045 	info.si_stime = nsec_to_clock_t(stime);
2046 
2047  	info.si_code = why;
2048  	switch (why) {
2049  	case CLD_CONTINUED:
2050  		info.si_status = SIGCONT;
2051  		break;
2052  	case CLD_STOPPED:
2053  		info.si_status = tsk->signal->group_exit_code & 0x7f;
2054  		break;
2055  	case CLD_TRAPPED:
2056  		info.si_status = tsk->exit_code & 0x7f;
2057  		break;
2058  	default:
2059  		BUG();
2060  	}
2061 
2062 	sighand = parent->sighand;
2063 	spin_lock_irqsave(&sighand->siglock, flags);
2064 	if (sighand->action[SIGCHLD-1].sa.sa_handler != SIG_IGN &&
2065 	    !(sighand->action[SIGCHLD-1].sa.sa_flags & SA_NOCLDSTOP))
2066 		__group_send_sig_info(SIGCHLD, &info, parent);
2067 	/*
2068 	 * Even if SIGCHLD is not generated, we must wake up wait4 calls.
2069 	 */
2070 	__wake_up_parent(tsk, parent);
2071 	spin_unlock_irqrestore(&sighand->siglock, flags);
2072 }
2073 
2074 static inline bool may_ptrace_stop(void)
2075 {
2076 	if (!likely(current->ptrace))
2077 		return false;
2078 	/*
2079 	 * Are we in the middle of do_coredump?
2080 	 * If so and our tracer is also part of the coredump stopping
2081 	 * is a deadlock situation, and pointless because our tracer
2082 	 * is dead so don't allow us to stop.
2083 	 * If SIGKILL was already sent before the caller unlocked
2084 	 * ->siglock we must see ->core_state != NULL. Otherwise it
2085 	 * is safe to enter schedule().
2086 	 *
2087 	 * This is almost outdated, a task with the pending SIGKILL can't
2088 	 * block in TASK_TRACED. But PTRACE_EVENT_EXIT can be reported
2089 	 * after SIGKILL was already dequeued.
2090 	 */
2091 	if (unlikely(current->mm->core_state) &&
2092 	    unlikely(current->mm == current->parent->mm))
2093 		return false;
2094 
2095 	return true;
2096 }
2097 
2098 /*
2099  * Return non-zero if there is a SIGKILL that should be waking us up.
2100  * Called with the siglock held.
2101  */
2102 static bool sigkill_pending(struct task_struct *tsk)
2103 {
2104 	return sigismember(&tsk->pending.signal, SIGKILL) ||
2105 	       sigismember(&tsk->signal->shared_pending.signal, SIGKILL);
2106 }
2107 
2108 /*
2109  * This must be called with current->sighand->siglock held.
2110  *
2111  * This should be the path for all ptrace stops.
2112  * We always set current->last_siginfo while stopped here.
2113  * That makes it a way to test a stopped process for
2114  * being ptrace-stopped vs being job-control-stopped.
2115  *
2116  * If we actually decide not to stop at all because the tracer
2117  * is gone, we keep current->exit_code unless clear_code.
2118  */
2119 static void ptrace_stop(int exit_code, int why, int clear_code, kernel_siginfo_t *info)
2120 	__releases(&current->sighand->siglock)
2121 	__acquires(&current->sighand->siglock)
2122 {
2123 	bool gstop_done = false;
2124 
2125 	if (arch_ptrace_stop_needed(exit_code, info)) {
2126 		/*
2127 		 * The arch code has something special to do before a
2128 		 * ptrace stop.  This is allowed to block, e.g. for faults
2129 		 * on user stack pages.  We can't keep the siglock while
2130 		 * calling arch_ptrace_stop, so we must release it now.
2131 		 * To preserve proper semantics, we must do this before
2132 		 * any signal bookkeeping like checking group_stop_count.
2133 		 * Meanwhile, a SIGKILL could come in before we retake the
2134 		 * siglock.  That must prevent us from sleeping in TASK_TRACED.
2135 		 * So after regaining the lock, we must check for SIGKILL.
2136 		 */
2137 		spin_unlock_irq(&current->sighand->siglock);
2138 		arch_ptrace_stop(exit_code, info);
2139 		spin_lock_irq(&current->sighand->siglock);
2140 		if (sigkill_pending(current))
2141 			return;
2142 	}
2143 
2144 	set_special_state(TASK_TRACED);
2145 
2146 	/*
2147 	 * We're committing to trapping.  TRACED should be visible before
2148 	 * TRAPPING is cleared; otherwise, the tracer might fail do_wait().
2149 	 * Also, transition to TRACED and updates to ->jobctl should be
2150 	 * atomic with respect to siglock and should be done after the arch
2151 	 * hook as siglock is released and regrabbed across it.
2152 	 *
2153 	 *     TRACER				    TRACEE
2154 	 *
2155 	 *     ptrace_attach()
2156 	 * [L]   wait_on_bit(JOBCTL_TRAPPING)	[S] set_special_state(TRACED)
2157 	 *     do_wait()
2158 	 *       set_current_state()                smp_wmb();
2159 	 *       ptrace_do_wait()
2160 	 *         wait_task_stopped()
2161 	 *           task_stopped_code()
2162 	 * [L]         task_is_traced()		[S] task_clear_jobctl_trapping();
2163 	 */
2164 	smp_wmb();
2165 
2166 	current->last_siginfo = info;
2167 	current->exit_code = exit_code;
2168 
2169 	/*
2170 	 * If @why is CLD_STOPPED, we're trapping to participate in a group
2171 	 * stop.  Do the bookkeeping.  Note that if SIGCONT was delievered
2172 	 * across siglock relocks since INTERRUPT was scheduled, PENDING
2173 	 * could be clear now.  We act as if SIGCONT is received after
2174 	 * TASK_TRACED is entered - ignore it.
2175 	 */
2176 	if (why == CLD_STOPPED && (current->jobctl & JOBCTL_STOP_PENDING))
2177 		gstop_done = task_participate_group_stop(current);
2178 
2179 	/* any trap clears pending STOP trap, STOP trap clears NOTIFY */
2180 	task_clear_jobctl_pending(current, JOBCTL_TRAP_STOP);
2181 	if (info && info->si_code >> 8 == PTRACE_EVENT_STOP)
2182 		task_clear_jobctl_pending(current, JOBCTL_TRAP_NOTIFY);
2183 
2184 	/* entering a trap, clear TRAPPING */
2185 	task_clear_jobctl_trapping(current);
2186 
2187 	spin_unlock_irq(&current->sighand->siglock);
2188 	read_lock(&tasklist_lock);
2189 	if (may_ptrace_stop()) {
2190 		/*
2191 		 * Notify parents of the stop.
2192 		 *
2193 		 * While ptraced, there are two parents - the ptracer and
2194 		 * the real_parent of the group_leader.  The ptracer should
2195 		 * know about every stop while the real parent is only
2196 		 * interested in the completion of group stop.  The states
2197 		 * for the two don't interact with each other.  Notify
2198 		 * separately unless they're gonna be duplicates.
2199 		 */
2200 		do_notify_parent_cldstop(current, true, why);
2201 		if (gstop_done && ptrace_reparented(current))
2202 			do_notify_parent_cldstop(current, false, why);
2203 
2204 		/*
2205 		 * Don't want to allow preemption here, because
2206 		 * sys_ptrace() needs this task to be inactive.
2207 		 *
2208 		 * XXX: implement read_unlock_no_resched().
2209 		 */
2210 		preempt_disable();
2211 		read_unlock(&tasklist_lock);
2212 		cgroup_enter_frozen();
2213 		preempt_enable_no_resched();
2214 		freezable_schedule();
2215 		cgroup_leave_frozen(true);
2216 	} else {
2217 		/*
2218 		 * By the time we got the lock, our tracer went away.
2219 		 * Don't drop the lock yet, another tracer may come.
2220 		 *
2221 		 * If @gstop_done, the ptracer went away between group stop
2222 		 * completion and here.  During detach, it would have set
2223 		 * JOBCTL_STOP_PENDING on us and we'll re-enter
2224 		 * TASK_STOPPED in do_signal_stop() on return, so notifying
2225 		 * the real parent of the group stop completion is enough.
2226 		 */
2227 		if (gstop_done)
2228 			do_notify_parent_cldstop(current, false, why);
2229 
2230 		/* tasklist protects us from ptrace_freeze_traced() */
2231 		__set_current_state(TASK_RUNNING);
2232 		if (clear_code)
2233 			current->exit_code = 0;
2234 		read_unlock(&tasklist_lock);
2235 	}
2236 
2237 	/*
2238 	 * We are back.  Now reacquire the siglock before touching
2239 	 * last_siginfo, so that we are sure to have synchronized with
2240 	 * any signal-sending on another CPU that wants to examine it.
2241 	 */
2242 	spin_lock_irq(&current->sighand->siglock);
2243 	current->last_siginfo = NULL;
2244 
2245 	/* LISTENING can be set only during STOP traps, clear it */
2246 	current->jobctl &= ~JOBCTL_LISTENING;
2247 
2248 	/*
2249 	 * Queued signals ignored us while we were stopped for tracing.
2250 	 * So check for any that we should take before resuming user mode.
2251 	 * This sets TIF_SIGPENDING, but never clears it.
2252 	 */
2253 	recalc_sigpending_tsk(current);
2254 }
2255 
2256 static void ptrace_do_notify(int signr, int exit_code, int why)
2257 {
2258 	kernel_siginfo_t info;
2259 
2260 	clear_siginfo(&info);
2261 	info.si_signo = signr;
2262 	info.si_code = exit_code;
2263 	info.si_pid = task_pid_vnr(current);
2264 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
2265 
2266 	/* Let the debugger run.  */
2267 	ptrace_stop(exit_code, why, 1, &info);
2268 }
2269 
2270 void ptrace_notify(int exit_code)
2271 {
2272 	BUG_ON((exit_code & (0x7f | ~0xffff)) != SIGTRAP);
2273 	if (unlikely(current->task_works))
2274 		task_work_run();
2275 
2276 	spin_lock_irq(&current->sighand->siglock);
2277 	ptrace_do_notify(SIGTRAP, exit_code, CLD_TRAPPED);
2278 	spin_unlock_irq(&current->sighand->siglock);
2279 }
2280 
2281 /**
2282  * do_signal_stop - handle group stop for SIGSTOP and other stop signals
2283  * @signr: signr causing group stop if initiating
2284  *
2285  * If %JOBCTL_STOP_PENDING is not set yet, initiate group stop with @signr
2286  * and participate in it.  If already set, participate in the existing
2287  * group stop.  If participated in a group stop (and thus slept), %true is
2288  * returned with siglock released.
2289  *
2290  * If ptraced, this function doesn't handle stop itself.  Instead,
2291  * %JOBCTL_TRAP_STOP is scheduled and %false is returned with siglock
2292  * untouched.  The caller must ensure that INTERRUPT trap handling takes
2293  * places afterwards.
2294  *
2295  * CONTEXT:
2296  * Must be called with @current->sighand->siglock held, which is released
2297  * on %true return.
2298  *
2299  * RETURNS:
2300  * %false if group stop is already cancelled or ptrace trap is scheduled.
2301  * %true if participated in group stop.
2302  */
2303 static bool do_signal_stop(int signr)
2304 	__releases(&current->sighand->siglock)
2305 {
2306 	struct signal_struct *sig = current->signal;
2307 
2308 	if (!(current->jobctl & JOBCTL_STOP_PENDING)) {
2309 		unsigned long gstop = JOBCTL_STOP_PENDING | JOBCTL_STOP_CONSUME;
2310 		struct task_struct *t;
2311 
2312 		/* signr will be recorded in task->jobctl for retries */
2313 		WARN_ON_ONCE(signr & ~JOBCTL_STOP_SIGMASK);
2314 
2315 		if (!likely(current->jobctl & JOBCTL_STOP_DEQUEUED) ||
2316 		    unlikely(signal_group_exit(sig)))
2317 			return false;
2318 		/*
2319 		 * There is no group stop already in progress.  We must
2320 		 * initiate one now.
2321 		 *
2322 		 * While ptraced, a task may be resumed while group stop is
2323 		 * still in effect and then receive a stop signal and
2324 		 * initiate another group stop.  This deviates from the
2325 		 * usual behavior as two consecutive stop signals can't
2326 		 * cause two group stops when !ptraced.  That is why we
2327 		 * also check !task_is_stopped(t) below.
2328 		 *
2329 		 * The condition can be distinguished by testing whether
2330 		 * SIGNAL_STOP_STOPPED is already set.  Don't generate
2331 		 * group_exit_code in such case.
2332 		 *
2333 		 * This is not necessary for SIGNAL_STOP_CONTINUED because
2334 		 * an intervening stop signal is required to cause two
2335 		 * continued events regardless of ptrace.
2336 		 */
2337 		if (!(sig->flags & SIGNAL_STOP_STOPPED))
2338 			sig->group_exit_code = signr;
2339 
2340 		sig->group_stop_count = 0;
2341 
2342 		if (task_set_jobctl_pending(current, signr | gstop))
2343 			sig->group_stop_count++;
2344 
2345 		t = current;
2346 		while_each_thread(current, t) {
2347 			/*
2348 			 * Setting state to TASK_STOPPED for a group
2349 			 * stop is always done with the siglock held,
2350 			 * so this check has no races.
2351 			 */
2352 			if (!task_is_stopped(t) &&
2353 			    task_set_jobctl_pending(t, signr | gstop)) {
2354 				sig->group_stop_count++;
2355 				if (likely(!(t->ptrace & PT_SEIZED)))
2356 					signal_wake_up(t, 0);
2357 				else
2358 					ptrace_trap_notify(t);
2359 			}
2360 		}
2361 	}
2362 
2363 	if (likely(!current->ptrace)) {
2364 		int notify = 0;
2365 
2366 		/*
2367 		 * If there are no other threads in the group, or if there
2368 		 * is a group stop in progress and we are the last to stop,
2369 		 * report to the parent.
2370 		 */
2371 		if (task_participate_group_stop(current))
2372 			notify = CLD_STOPPED;
2373 
2374 		set_special_state(TASK_STOPPED);
2375 		spin_unlock_irq(&current->sighand->siglock);
2376 
2377 		/*
2378 		 * Notify the parent of the group stop completion.  Because
2379 		 * we're not holding either the siglock or tasklist_lock
2380 		 * here, ptracer may attach inbetween; however, this is for
2381 		 * group stop and should always be delivered to the real
2382 		 * parent of the group leader.  The new ptracer will get
2383 		 * its notification when this task transitions into
2384 		 * TASK_TRACED.
2385 		 */
2386 		if (notify) {
2387 			read_lock(&tasklist_lock);
2388 			do_notify_parent_cldstop(current, false, notify);
2389 			read_unlock(&tasklist_lock);
2390 		}
2391 
2392 		/* Now we don't run again until woken by SIGCONT or SIGKILL */
2393 		cgroup_enter_frozen();
2394 		freezable_schedule();
2395 		return true;
2396 	} else {
2397 		/*
2398 		 * While ptraced, group stop is handled by STOP trap.
2399 		 * Schedule it and let the caller deal with it.
2400 		 */
2401 		task_set_jobctl_pending(current, JOBCTL_TRAP_STOP);
2402 		return false;
2403 	}
2404 }
2405 
2406 /**
2407  * do_jobctl_trap - take care of ptrace jobctl traps
2408  *
2409  * When PT_SEIZED, it's used for both group stop and explicit
2410  * SEIZE/INTERRUPT traps.  Both generate PTRACE_EVENT_STOP trap with
2411  * accompanying siginfo.  If stopped, lower eight bits of exit_code contain
2412  * the stop signal; otherwise, %SIGTRAP.
2413  *
2414  * When !PT_SEIZED, it's used only for group stop trap with stop signal
2415  * number as exit_code and no siginfo.
2416  *
2417  * CONTEXT:
2418  * Must be called with @current->sighand->siglock held, which may be
2419  * released and re-acquired before returning with intervening sleep.
2420  */
2421 static void do_jobctl_trap(void)
2422 {
2423 	struct signal_struct *signal = current->signal;
2424 	int signr = current->jobctl & JOBCTL_STOP_SIGMASK;
2425 
2426 	if (current->ptrace & PT_SEIZED) {
2427 		if (!signal->group_stop_count &&
2428 		    !(signal->flags & SIGNAL_STOP_STOPPED))
2429 			signr = SIGTRAP;
2430 		WARN_ON_ONCE(!signr);
2431 		ptrace_do_notify(signr, signr | (PTRACE_EVENT_STOP << 8),
2432 				 CLD_STOPPED);
2433 	} else {
2434 		WARN_ON_ONCE(!signr);
2435 		ptrace_stop(signr, CLD_STOPPED, 0, NULL);
2436 		current->exit_code = 0;
2437 	}
2438 }
2439 
2440 /**
2441  * do_freezer_trap - handle the freezer jobctl trap
2442  *
2443  * Puts the task into frozen state, if only the task is not about to quit.
2444  * In this case it drops JOBCTL_TRAP_FREEZE.
2445  *
2446  * CONTEXT:
2447  * Must be called with @current->sighand->siglock held,
2448  * which is always released before returning.
2449  */
2450 static void do_freezer_trap(void)
2451 	__releases(&current->sighand->siglock)
2452 {
2453 	/*
2454 	 * If there are other trap bits pending except JOBCTL_TRAP_FREEZE,
2455 	 * let's make another loop to give it a chance to be handled.
2456 	 * In any case, we'll return back.
2457 	 */
2458 	if ((current->jobctl & (JOBCTL_PENDING_MASK | JOBCTL_TRAP_FREEZE)) !=
2459 	     JOBCTL_TRAP_FREEZE) {
2460 		spin_unlock_irq(&current->sighand->siglock);
2461 		return;
2462 	}
2463 
2464 	/*
2465 	 * Now we're sure that there is no pending fatal signal and no
2466 	 * pending traps. Clear TIF_SIGPENDING to not get out of schedule()
2467 	 * immediately (if there is a non-fatal signal pending), and
2468 	 * put the task into sleep.
2469 	 */
2470 	__set_current_state(TASK_INTERRUPTIBLE);
2471 	clear_thread_flag(TIF_SIGPENDING);
2472 	spin_unlock_irq(&current->sighand->siglock);
2473 	cgroup_enter_frozen();
2474 	freezable_schedule();
2475 }
2476 
2477 static int ptrace_signal(int signr, kernel_siginfo_t *info)
2478 {
2479 	/*
2480 	 * We do not check sig_kernel_stop(signr) but set this marker
2481 	 * unconditionally because we do not know whether debugger will
2482 	 * change signr. This flag has no meaning unless we are going
2483 	 * to stop after return from ptrace_stop(). In this case it will
2484 	 * be checked in do_signal_stop(), we should only stop if it was
2485 	 * not cleared by SIGCONT while we were sleeping. See also the
2486 	 * comment in dequeue_signal().
2487 	 */
2488 	current->jobctl |= JOBCTL_STOP_DEQUEUED;
2489 	ptrace_stop(signr, CLD_TRAPPED, 0, info);
2490 
2491 	/* We're back.  Did the debugger cancel the sig?  */
2492 	signr = current->exit_code;
2493 	if (signr == 0)
2494 		return signr;
2495 
2496 	current->exit_code = 0;
2497 
2498 	/*
2499 	 * Update the siginfo structure if the signal has
2500 	 * changed.  If the debugger wanted something
2501 	 * specific in the siginfo structure then it should
2502 	 * have updated *info via PTRACE_SETSIGINFO.
2503 	 */
2504 	if (signr != info->si_signo) {
2505 		clear_siginfo(info);
2506 		info->si_signo = signr;
2507 		info->si_errno = 0;
2508 		info->si_code = SI_USER;
2509 		rcu_read_lock();
2510 		info->si_pid = task_pid_vnr(current->parent);
2511 		info->si_uid = from_kuid_munged(current_user_ns(),
2512 						task_uid(current->parent));
2513 		rcu_read_unlock();
2514 	}
2515 
2516 	/* If the (new) signal is now blocked, requeue it.  */
2517 	if (sigismember(&current->blocked, signr)) {
2518 		send_signal(signr, info, current, PIDTYPE_PID);
2519 		signr = 0;
2520 	}
2521 
2522 	return signr;
2523 }
2524 
2525 static void hide_si_addr_tag_bits(struct ksignal *ksig)
2526 {
2527 	switch (siginfo_layout(ksig->sig, ksig->info.si_code)) {
2528 	case SIL_FAULT:
2529 	case SIL_FAULT_MCEERR:
2530 	case SIL_FAULT_BNDERR:
2531 	case SIL_FAULT_PKUERR:
2532 		ksig->info.si_addr = arch_untagged_si_addr(
2533 			ksig->info.si_addr, ksig->sig, ksig->info.si_code);
2534 		break;
2535 	case SIL_KILL:
2536 	case SIL_TIMER:
2537 	case SIL_POLL:
2538 	case SIL_CHLD:
2539 	case SIL_RT:
2540 	case SIL_SYS:
2541 		break;
2542 	}
2543 }
2544 
2545 bool get_signal(struct ksignal *ksig)
2546 {
2547 	struct sighand_struct *sighand = current->sighand;
2548 	struct signal_struct *signal = current->signal;
2549 	int signr;
2550 
2551 	if (unlikely(current->task_works))
2552 		task_work_run();
2553 
2554 	/*
2555 	 * For non-generic architectures, check for TIF_NOTIFY_SIGNAL so
2556 	 * that the arch handlers don't all have to do it. If we get here
2557 	 * without TIF_SIGPENDING, just exit after running signal work.
2558 	 */
2559 	if (!IS_ENABLED(CONFIG_GENERIC_ENTRY)) {
2560 		if (test_thread_flag(TIF_NOTIFY_SIGNAL))
2561 			tracehook_notify_signal();
2562 		if (!task_sigpending(current))
2563 			return false;
2564 	}
2565 
2566 	if (unlikely(uprobe_deny_signal()))
2567 		return false;
2568 
2569 	/*
2570 	 * Do this once, we can't return to user-mode if freezing() == T.
2571 	 * do_signal_stop() and ptrace_stop() do freezable_schedule() and
2572 	 * thus do not need another check after return.
2573 	 */
2574 	try_to_freeze();
2575 
2576 relock:
2577 	spin_lock_irq(&sighand->siglock);
2578 
2579 	/*
2580 	 * Every stopped thread goes here after wakeup. Check to see if
2581 	 * we should notify the parent, prepare_signal(SIGCONT) encodes
2582 	 * the CLD_ si_code into SIGNAL_CLD_MASK bits.
2583 	 */
2584 	if (unlikely(signal->flags & SIGNAL_CLD_MASK)) {
2585 		int why;
2586 
2587 		if (signal->flags & SIGNAL_CLD_CONTINUED)
2588 			why = CLD_CONTINUED;
2589 		else
2590 			why = CLD_STOPPED;
2591 
2592 		signal->flags &= ~SIGNAL_CLD_MASK;
2593 
2594 		spin_unlock_irq(&sighand->siglock);
2595 
2596 		/*
2597 		 * Notify the parent that we're continuing.  This event is
2598 		 * always per-process and doesn't make whole lot of sense
2599 		 * for ptracers, who shouldn't consume the state via
2600 		 * wait(2) either, but, for backward compatibility, notify
2601 		 * the ptracer of the group leader too unless it's gonna be
2602 		 * a duplicate.
2603 		 */
2604 		read_lock(&tasklist_lock);
2605 		do_notify_parent_cldstop(current, false, why);
2606 
2607 		if (ptrace_reparented(current->group_leader))
2608 			do_notify_parent_cldstop(current->group_leader,
2609 						true, why);
2610 		read_unlock(&tasklist_lock);
2611 
2612 		goto relock;
2613 	}
2614 
2615 	/* Has this task already been marked for death? */
2616 	if (signal_group_exit(signal)) {
2617 		ksig->info.si_signo = signr = SIGKILL;
2618 		sigdelset(&current->pending.signal, SIGKILL);
2619 		trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
2620 				&sighand->action[SIGKILL - 1]);
2621 		recalc_sigpending();
2622 		goto fatal;
2623 	}
2624 
2625 	for (;;) {
2626 		struct k_sigaction *ka;
2627 
2628 		if (unlikely(current->jobctl & JOBCTL_STOP_PENDING) &&
2629 		    do_signal_stop(0))
2630 			goto relock;
2631 
2632 		if (unlikely(current->jobctl &
2633 			     (JOBCTL_TRAP_MASK | JOBCTL_TRAP_FREEZE))) {
2634 			if (current->jobctl & JOBCTL_TRAP_MASK) {
2635 				do_jobctl_trap();
2636 				spin_unlock_irq(&sighand->siglock);
2637 			} else if (current->jobctl & JOBCTL_TRAP_FREEZE)
2638 				do_freezer_trap();
2639 
2640 			goto relock;
2641 		}
2642 
2643 		/*
2644 		 * If the task is leaving the frozen state, let's update
2645 		 * cgroup counters and reset the frozen bit.
2646 		 */
2647 		if (unlikely(cgroup_task_frozen(current))) {
2648 			spin_unlock_irq(&sighand->siglock);
2649 			cgroup_leave_frozen(false);
2650 			goto relock;
2651 		}
2652 
2653 		/*
2654 		 * Signals generated by the execution of an instruction
2655 		 * need to be delivered before any other pending signals
2656 		 * so that the instruction pointer in the signal stack
2657 		 * frame points to the faulting instruction.
2658 		 */
2659 		signr = dequeue_synchronous_signal(&ksig->info);
2660 		if (!signr)
2661 			signr = dequeue_signal(current, &current->blocked, &ksig->info);
2662 
2663 		if (!signr)
2664 			break; /* will return 0 */
2665 
2666 		if (unlikely(current->ptrace) && signr != SIGKILL) {
2667 			signr = ptrace_signal(signr, &ksig->info);
2668 			if (!signr)
2669 				continue;
2670 		}
2671 
2672 		ka = &sighand->action[signr-1];
2673 
2674 		/* Trace actually delivered signals. */
2675 		trace_signal_deliver(signr, &ksig->info, ka);
2676 
2677 		if (ka->sa.sa_handler == SIG_IGN) /* Do nothing.  */
2678 			continue;
2679 		if (ka->sa.sa_handler != SIG_DFL) {
2680 			/* Run the handler.  */
2681 			ksig->ka = *ka;
2682 
2683 			if (ka->sa.sa_flags & SA_ONESHOT)
2684 				ka->sa.sa_handler = SIG_DFL;
2685 
2686 			break; /* will return non-zero "signr" value */
2687 		}
2688 
2689 		/*
2690 		 * Now we are doing the default action for this signal.
2691 		 */
2692 		if (sig_kernel_ignore(signr)) /* Default is nothing. */
2693 			continue;
2694 
2695 		/*
2696 		 * Global init gets no signals it doesn't want.
2697 		 * Container-init gets no signals it doesn't want from same
2698 		 * container.
2699 		 *
2700 		 * Note that if global/container-init sees a sig_kernel_only()
2701 		 * signal here, the signal must have been generated internally
2702 		 * or must have come from an ancestor namespace. In either
2703 		 * case, the signal cannot be dropped.
2704 		 */
2705 		if (unlikely(signal->flags & SIGNAL_UNKILLABLE) &&
2706 				!sig_kernel_only(signr))
2707 			continue;
2708 
2709 		if (sig_kernel_stop(signr)) {
2710 			/*
2711 			 * The default action is to stop all threads in
2712 			 * the thread group.  The job control signals
2713 			 * do nothing in an orphaned pgrp, but SIGSTOP
2714 			 * always works.  Note that siglock needs to be
2715 			 * dropped during the call to is_orphaned_pgrp()
2716 			 * because of lock ordering with tasklist_lock.
2717 			 * This allows an intervening SIGCONT to be posted.
2718 			 * We need to check for that and bail out if necessary.
2719 			 */
2720 			if (signr != SIGSTOP) {
2721 				spin_unlock_irq(&sighand->siglock);
2722 
2723 				/* signals can be posted during this window */
2724 
2725 				if (is_current_pgrp_orphaned())
2726 					goto relock;
2727 
2728 				spin_lock_irq(&sighand->siglock);
2729 			}
2730 
2731 			if (likely(do_signal_stop(ksig->info.si_signo))) {
2732 				/* It released the siglock.  */
2733 				goto relock;
2734 			}
2735 
2736 			/*
2737 			 * We didn't actually stop, due to a race
2738 			 * with SIGCONT or something like that.
2739 			 */
2740 			continue;
2741 		}
2742 
2743 	fatal:
2744 		spin_unlock_irq(&sighand->siglock);
2745 		if (unlikely(cgroup_task_frozen(current)))
2746 			cgroup_leave_frozen(true);
2747 
2748 		/*
2749 		 * Anything else is fatal, maybe with a core dump.
2750 		 */
2751 		current->flags |= PF_SIGNALED;
2752 
2753 		if (sig_kernel_coredump(signr)) {
2754 			if (print_fatal_signals)
2755 				print_fatal_signal(ksig->info.si_signo);
2756 			proc_coredump_connector(current);
2757 			/*
2758 			 * If it was able to dump core, this kills all
2759 			 * other threads in the group and synchronizes with
2760 			 * their demise.  If we lost the race with another
2761 			 * thread getting here, it set group_exit_code
2762 			 * first and our do_group_exit call below will use
2763 			 * that value and ignore the one we pass it.
2764 			 */
2765 			do_coredump(&ksig->info);
2766 		}
2767 
2768 		/*
2769 		 * PF_IO_WORKER threads will catch and exit on fatal signals
2770 		 * themselves. They have cleanup that must be performed, so
2771 		 * we cannot call do_exit() on their behalf.
2772 		 */
2773 		if (current->flags & PF_IO_WORKER)
2774 			goto out;
2775 
2776 		/*
2777 		 * Death signals, no core dump.
2778 		 */
2779 		do_group_exit(ksig->info.si_signo);
2780 		/* NOTREACHED */
2781 	}
2782 	spin_unlock_irq(&sighand->siglock);
2783 out:
2784 	ksig->sig = signr;
2785 
2786 	if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
2787 		hide_si_addr_tag_bits(ksig);
2788 
2789 	return ksig->sig > 0;
2790 }
2791 
2792 /**
2793  * signal_delivered -
2794  * @ksig:		kernel signal struct
2795  * @stepping:		nonzero if debugger single-step or block-step in use
2796  *
2797  * This function should be called when a signal has successfully been
2798  * delivered. It updates the blocked signals accordingly (@ksig->ka.sa.sa_mask
2799  * is always blocked, and the signal itself is blocked unless %SA_NODEFER
2800  * is set in @ksig->ka.sa.sa_flags.  Tracing is notified.
2801  */
2802 static void signal_delivered(struct ksignal *ksig, int stepping)
2803 {
2804 	sigset_t blocked;
2805 
2806 	/* A signal was successfully delivered, and the
2807 	   saved sigmask was stored on the signal frame,
2808 	   and will be restored by sigreturn.  So we can
2809 	   simply clear the restore sigmask flag.  */
2810 	clear_restore_sigmask();
2811 
2812 	sigorsets(&blocked, &current->blocked, &ksig->ka.sa.sa_mask);
2813 	if (!(ksig->ka.sa.sa_flags & SA_NODEFER))
2814 		sigaddset(&blocked, ksig->sig);
2815 	set_current_blocked(&blocked);
2816 	tracehook_signal_handler(stepping);
2817 }
2818 
2819 void signal_setup_done(int failed, struct ksignal *ksig, int stepping)
2820 {
2821 	if (failed)
2822 		force_sigsegv(ksig->sig);
2823 	else
2824 		signal_delivered(ksig, stepping);
2825 }
2826 
2827 /*
2828  * It could be that complete_signal() picked us to notify about the
2829  * group-wide signal. Other threads should be notified now to take
2830  * the shared signals in @which since we will not.
2831  */
2832 static void retarget_shared_pending(struct task_struct *tsk, sigset_t *which)
2833 {
2834 	sigset_t retarget;
2835 	struct task_struct *t;
2836 
2837 	sigandsets(&retarget, &tsk->signal->shared_pending.signal, which);
2838 	if (sigisemptyset(&retarget))
2839 		return;
2840 
2841 	t = tsk;
2842 	while_each_thread(tsk, t) {
2843 		if (t->flags & PF_EXITING)
2844 			continue;
2845 
2846 		if (!has_pending_signals(&retarget, &t->blocked))
2847 			continue;
2848 		/* Remove the signals this thread can handle. */
2849 		sigandsets(&retarget, &retarget, &t->blocked);
2850 
2851 		if (!task_sigpending(t))
2852 			signal_wake_up(t, 0);
2853 
2854 		if (sigisemptyset(&retarget))
2855 			break;
2856 	}
2857 }
2858 
2859 void exit_signals(struct task_struct *tsk)
2860 {
2861 	int group_stop = 0;
2862 	sigset_t unblocked;
2863 
2864 	/*
2865 	 * @tsk is about to have PF_EXITING set - lock out users which
2866 	 * expect stable threadgroup.
2867 	 */
2868 	cgroup_threadgroup_change_begin(tsk);
2869 
2870 	if (thread_group_empty(tsk) || signal_group_exit(tsk->signal)) {
2871 		tsk->flags |= PF_EXITING;
2872 		cgroup_threadgroup_change_end(tsk);
2873 		return;
2874 	}
2875 
2876 	spin_lock_irq(&tsk->sighand->siglock);
2877 	/*
2878 	 * From now this task is not visible for group-wide signals,
2879 	 * see wants_signal(), do_signal_stop().
2880 	 */
2881 	tsk->flags |= PF_EXITING;
2882 
2883 	cgroup_threadgroup_change_end(tsk);
2884 
2885 	if (!task_sigpending(tsk))
2886 		goto out;
2887 
2888 	unblocked = tsk->blocked;
2889 	signotset(&unblocked);
2890 	retarget_shared_pending(tsk, &unblocked);
2891 
2892 	if (unlikely(tsk->jobctl & JOBCTL_STOP_PENDING) &&
2893 	    task_participate_group_stop(tsk))
2894 		group_stop = CLD_STOPPED;
2895 out:
2896 	spin_unlock_irq(&tsk->sighand->siglock);
2897 
2898 	/*
2899 	 * If group stop has completed, deliver the notification.  This
2900 	 * should always go to the real parent of the group leader.
2901 	 */
2902 	if (unlikely(group_stop)) {
2903 		read_lock(&tasklist_lock);
2904 		do_notify_parent_cldstop(tsk, false, group_stop);
2905 		read_unlock(&tasklist_lock);
2906 	}
2907 }
2908 
2909 /*
2910  * System call entry points.
2911  */
2912 
2913 /**
2914  *  sys_restart_syscall - restart a system call
2915  */
2916 SYSCALL_DEFINE0(restart_syscall)
2917 {
2918 	struct restart_block *restart = &current->restart_block;
2919 	return restart->fn(restart);
2920 }
2921 
2922 long do_no_restart_syscall(struct restart_block *param)
2923 {
2924 	return -EINTR;
2925 }
2926 
2927 static void __set_task_blocked(struct task_struct *tsk, const sigset_t *newset)
2928 {
2929 	if (task_sigpending(tsk) && !thread_group_empty(tsk)) {
2930 		sigset_t newblocked;
2931 		/* A set of now blocked but previously unblocked signals. */
2932 		sigandnsets(&newblocked, newset, &current->blocked);
2933 		retarget_shared_pending(tsk, &newblocked);
2934 	}
2935 	tsk->blocked = *newset;
2936 	recalc_sigpending();
2937 }
2938 
2939 /**
2940  * set_current_blocked - change current->blocked mask
2941  * @newset: new mask
2942  *
2943  * It is wrong to change ->blocked directly, this helper should be used
2944  * to ensure the process can't miss a shared signal we are going to block.
2945  */
2946 void set_current_blocked(sigset_t *newset)
2947 {
2948 	sigdelsetmask(newset, sigmask(SIGKILL) | sigmask(SIGSTOP));
2949 	__set_current_blocked(newset);
2950 }
2951 
2952 void __set_current_blocked(const sigset_t *newset)
2953 {
2954 	struct task_struct *tsk = current;
2955 
2956 	/*
2957 	 * In case the signal mask hasn't changed, there is nothing we need
2958 	 * to do. The current->blocked shouldn't be modified by other task.
2959 	 */
2960 	if (sigequalsets(&tsk->blocked, newset))
2961 		return;
2962 
2963 	spin_lock_irq(&tsk->sighand->siglock);
2964 	__set_task_blocked(tsk, newset);
2965 	spin_unlock_irq(&tsk->sighand->siglock);
2966 }
2967 
2968 /*
2969  * This is also useful for kernel threads that want to temporarily
2970  * (or permanently) block certain signals.
2971  *
2972  * NOTE! Unlike the user-mode sys_sigprocmask(), the kernel
2973  * interface happily blocks "unblockable" signals like SIGKILL
2974  * and friends.
2975  */
2976 int sigprocmask(int how, sigset_t *set, sigset_t *oldset)
2977 {
2978 	struct task_struct *tsk = current;
2979 	sigset_t newset;
2980 
2981 	/* Lockless, only current can change ->blocked, never from irq */
2982 	if (oldset)
2983 		*oldset = tsk->blocked;
2984 
2985 	switch (how) {
2986 	case SIG_BLOCK:
2987 		sigorsets(&newset, &tsk->blocked, set);
2988 		break;
2989 	case SIG_UNBLOCK:
2990 		sigandnsets(&newset, &tsk->blocked, set);
2991 		break;
2992 	case SIG_SETMASK:
2993 		newset = *set;
2994 		break;
2995 	default:
2996 		return -EINVAL;
2997 	}
2998 
2999 	__set_current_blocked(&newset);
3000 	return 0;
3001 }
3002 EXPORT_SYMBOL(sigprocmask);
3003 
3004 /*
3005  * The api helps set app-provided sigmasks.
3006  *
3007  * This is useful for syscalls such as ppoll, pselect, io_pgetevents and
3008  * epoll_pwait where a new sigmask is passed from userland for the syscalls.
3009  *
3010  * Note that it does set_restore_sigmask() in advance, so it must be always
3011  * paired with restore_saved_sigmask_unless() before return from syscall.
3012  */
3013 int set_user_sigmask(const sigset_t __user *umask, size_t sigsetsize)
3014 {
3015 	sigset_t kmask;
3016 
3017 	if (!umask)
3018 		return 0;
3019 	if (sigsetsize != sizeof(sigset_t))
3020 		return -EINVAL;
3021 	if (copy_from_user(&kmask, umask, sizeof(sigset_t)))
3022 		return -EFAULT;
3023 
3024 	set_restore_sigmask();
3025 	current->saved_sigmask = current->blocked;
3026 	set_current_blocked(&kmask);
3027 
3028 	return 0;
3029 }
3030 
3031 #ifdef CONFIG_COMPAT
3032 int set_compat_user_sigmask(const compat_sigset_t __user *umask,
3033 			    size_t sigsetsize)
3034 {
3035 	sigset_t kmask;
3036 
3037 	if (!umask)
3038 		return 0;
3039 	if (sigsetsize != sizeof(compat_sigset_t))
3040 		return -EINVAL;
3041 	if (get_compat_sigset(&kmask, umask))
3042 		return -EFAULT;
3043 
3044 	set_restore_sigmask();
3045 	current->saved_sigmask = current->blocked;
3046 	set_current_blocked(&kmask);
3047 
3048 	return 0;
3049 }
3050 #endif
3051 
3052 /**
3053  *  sys_rt_sigprocmask - change the list of currently blocked signals
3054  *  @how: whether to add, remove, or set signals
3055  *  @nset: stores pending signals
3056  *  @oset: previous value of signal mask if non-null
3057  *  @sigsetsize: size of sigset_t type
3058  */
3059 SYSCALL_DEFINE4(rt_sigprocmask, int, how, sigset_t __user *, nset,
3060 		sigset_t __user *, oset, size_t, sigsetsize)
3061 {
3062 	sigset_t old_set, new_set;
3063 	int error;
3064 
3065 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3066 	if (sigsetsize != sizeof(sigset_t))
3067 		return -EINVAL;
3068 
3069 	old_set = current->blocked;
3070 
3071 	if (nset) {
3072 		if (copy_from_user(&new_set, nset, sizeof(sigset_t)))
3073 			return -EFAULT;
3074 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3075 
3076 		error = sigprocmask(how, &new_set, NULL);
3077 		if (error)
3078 			return error;
3079 	}
3080 
3081 	if (oset) {
3082 		if (copy_to_user(oset, &old_set, sizeof(sigset_t)))
3083 			return -EFAULT;
3084 	}
3085 
3086 	return 0;
3087 }
3088 
3089 #ifdef CONFIG_COMPAT
3090 COMPAT_SYSCALL_DEFINE4(rt_sigprocmask, int, how, compat_sigset_t __user *, nset,
3091 		compat_sigset_t __user *, oset, compat_size_t, sigsetsize)
3092 {
3093 	sigset_t old_set = current->blocked;
3094 
3095 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3096 	if (sigsetsize != sizeof(sigset_t))
3097 		return -EINVAL;
3098 
3099 	if (nset) {
3100 		sigset_t new_set;
3101 		int error;
3102 		if (get_compat_sigset(&new_set, nset))
3103 			return -EFAULT;
3104 		sigdelsetmask(&new_set, sigmask(SIGKILL)|sigmask(SIGSTOP));
3105 
3106 		error = sigprocmask(how, &new_set, NULL);
3107 		if (error)
3108 			return error;
3109 	}
3110 	return oset ? put_compat_sigset(oset, &old_set, sizeof(*oset)) : 0;
3111 }
3112 #endif
3113 
3114 static void do_sigpending(sigset_t *set)
3115 {
3116 	spin_lock_irq(&current->sighand->siglock);
3117 	sigorsets(set, &current->pending.signal,
3118 		  &current->signal->shared_pending.signal);
3119 	spin_unlock_irq(&current->sighand->siglock);
3120 
3121 	/* Outside the lock because only this thread touches it.  */
3122 	sigandsets(set, &current->blocked, set);
3123 }
3124 
3125 /**
3126  *  sys_rt_sigpending - examine a pending signal that has been raised
3127  *			while blocked
3128  *  @uset: stores pending signals
3129  *  @sigsetsize: size of sigset_t type or larger
3130  */
3131 SYSCALL_DEFINE2(rt_sigpending, sigset_t __user *, uset, size_t, sigsetsize)
3132 {
3133 	sigset_t set;
3134 
3135 	if (sigsetsize > sizeof(*uset))
3136 		return -EINVAL;
3137 
3138 	do_sigpending(&set);
3139 
3140 	if (copy_to_user(uset, &set, sigsetsize))
3141 		return -EFAULT;
3142 
3143 	return 0;
3144 }
3145 
3146 #ifdef CONFIG_COMPAT
3147 COMPAT_SYSCALL_DEFINE2(rt_sigpending, compat_sigset_t __user *, uset,
3148 		compat_size_t, sigsetsize)
3149 {
3150 	sigset_t set;
3151 
3152 	if (sigsetsize > sizeof(*uset))
3153 		return -EINVAL;
3154 
3155 	do_sigpending(&set);
3156 
3157 	return put_compat_sigset(uset, &set, sigsetsize);
3158 }
3159 #endif
3160 
3161 static const struct {
3162 	unsigned char limit, layout;
3163 } sig_sicodes[] = {
3164 	[SIGILL]  = { NSIGILL,  SIL_FAULT },
3165 	[SIGFPE]  = { NSIGFPE,  SIL_FAULT },
3166 	[SIGSEGV] = { NSIGSEGV, SIL_FAULT },
3167 	[SIGBUS]  = { NSIGBUS,  SIL_FAULT },
3168 	[SIGTRAP] = { NSIGTRAP, SIL_FAULT },
3169 #if defined(SIGEMT)
3170 	[SIGEMT]  = { NSIGEMT,  SIL_FAULT },
3171 #endif
3172 	[SIGCHLD] = { NSIGCHLD, SIL_CHLD },
3173 	[SIGPOLL] = { NSIGPOLL, SIL_POLL },
3174 	[SIGSYS]  = { NSIGSYS,  SIL_SYS },
3175 };
3176 
3177 static bool known_siginfo_layout(unsigned sig, int si_code)
3178 {
3179 	if (si_code == SI_KERNEL)
3180 		return true;
3181 	else if ((si_code > SI_USER)) {
3182 		if (sig_specific_sicodes(sig)) {
3183 			if (si_code <= sig_sicodes[sig].limit)
3184 				return true;
3185 		}
3186 		else if (si_code <= NSIGPOLL)
3187 			return true;
3188 	}
3189 	else if (si_code >= SI_DETHREAD)
3190 		return true;
3191 	else if (si_code == SI_ASYNCNL)
3192 		return true;
3193 	return false;
3194 }
3195 
3196 enum siginfo_layout siginfo_layout(unsigned sig, int si_code)
3197 {
3198 	enum siginfo_layout layout = SIL_KILL;
3199 	if ((si_code > SI_USER) && (si_code < SI_KERNEL)) {
3200 		if ((sig < ARRAY_SIZE(sig_sicodes)) &&
3201 		    (si_code <= sig_sicodes[sig].limit)) {
3202 			layout = sig_sicodes[sig].layout;
3203 			/* Handle the exceptions */
3204 			if ((sig == SIGBUS) &&
3205 			    (si_code >= BUS_MCEERR_AR) && (si_code <= BUS_MCEERR_AO))
3206 				layout = SIL_FAULT_MCEERR;
3207 			else if ((sig == SIGSEGV) && (si_code == SEGV_BNDERR))
3208 				layout = SIL_FAULT_BNDERR;
3209 #ifdef SEGV_PKUERR
3210 			else if ((sig == SIGSEGV) && (si_code == SEGV_PKUERR))
3211 				layout = SIL_FAULT_PKUERR;
3212 #endif
3213 		}
3214 		else if (si_code <= NSIGPOLL)
3215 			layout = SIL_POLL;
3216 	} else {
3217 		if (si_code == SI_TIMER)
3218 			layout = SIL_TIMER;
3219 		else if (si_code == SI_SIGIO)
3220 			layout = SIL_POLL;
3221 		else if (si_code < 0)
3222 			layout = SIL_RT;
3223 	}
3224 	return layout;
3225 }
3226 
3227 static inline char __user *si_expansion(const siginfo_t __user *info)
3228 {
3229 	return ((char __user *)info) + sizeof(struct kernel_siginfo);
3230 }
3231 
3232 int copy_siginfo_to_user(siginfo_t __user *to, const kernel_siginfo_t *from)
3233 {
3234 	char __user *expansion = si_expansion(to);
3235 	if (copy_to_user(to, from , sizeof(struct kernel_siginfo)))
3236 		return -EFAULT;
3237 	if (clear_user(expansion, SI_EXPANSION_SIZE))
3238 		return -EFAULT;
3239 	return 0;
3240 }
3241 
3242 static int post_copy_siginfo_from_user(kernel_siginfo_t *info,
3243 				       const siginfo_t __user *from)
3244 {
3245 	if (unlikely(!known_siginfo_layout(info->si_signo, info->si_code))) {
3246 		char __user *expansion = si_expansion(from);
3247 		char buf[SI_EXPANSION_SIZE];
3248 		int i;
3249 		/*
3250 		 * An unknown si_code might need more than
3251 		 * sizeof(struct kernel_siginfo) bytes.  Verify all of the
3252 		 * extra bytes are 0.  This guarantees copy_siginfo_to_user
3253 		 * will return this data to userspace exactly.
3254 		 */
3255 		if (copy_from_user(&buf, expansion, SI_EXPANSION_SIZE))
3256 			return -EFAULT;
3257 		for (i = 0; i < SI_EXPANSION_SIZE; i++) {
3258 			if (buf[i] != 0)
3259 				return -E2BIG;
3260 		}
3261 	}
3262 	return 0;
3263 }
3264 
3265 static int __copy_siginfo_from_user(int signo, kernel_siginfo_t *to,
3266 				    const siginfo_t __user *from)
3267 {
3268 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3269 		return -EFAULT;
3270 	to->si_signo = signo;
3271 	return post_copy_siginfo_from_user(to, from);
3272 }
3273 
3274 int copy_siginfo_from_user(kernel_siginfo_t *to, const siginfo_t __user *from)
3275 {
3276 	if (copy_from_user(to, from, sizeof(struct kernel_siginfo)))
3277 		return -EFAULT;
3278 	return post_copy_siginfo_from_user(to, from);
3279 }
3280 
3281 #ifdef CONFIG_COMPAT
3282 /**
3283  * copy_siginfo_to_external32 - copy a kernel siginfo into a compat user siginfo
3284  * @to: compat siginfo destination
3285  * @from: kernel siginfo source
3286  *
3287  * Note: This function does not work properly for the SIGCHLD on x32, but
3288  * fortunately it doesn't have to.  The only valid callers for this function are
3289  * copy_siginfo_to_user32, which is overriden for x32 and the coredump code.
3290  * The latter does not care because SIGCHLD will never cause a coredump.
3291  */
3292 void copy_siginfo_to_external32(struct compat_siginfo *to,
3293 		const struct kernel_siginfo *from)
3294 {
3295 	memset(to, 0, sizeof(*to));
3296 
3297 	to->si_signo = from->si_signo;
3298 	to->si_errno = from->si_errno;
3299 	to->si_code  = from->si_code;
3300 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3301 	case SIL_KILL:
3302 		to->si_pid = from->si_pid;
3303 		to->si_uid = from->si_uid;
3304 		break;
3305 	case SIL_TIMER:
3306 		to->si_tid     = from->si_tid;
3307 		to->si_overrun = from->si_overrun;
3308 		to->si_int     = from->si_int;
3309 		break;
3310 	case SIL_POLL:
3311 		to->si_band = from->si_band;
3312 		to->si_fd   = from->si_fd;
3313 		break;
3314 	case SIL_FAULT:
3315 		to->si_addr = ptr_to_compat(from->si_addr);
3316 #ifdef __ARCH_SI_TRAPNO
3317 		to->si_trapno = from->si_trapno;
3318 #endif
3319 		break;
3320 	case SIL_FAULT_MCEERR:
3321 		to->si_addr = ptr_to_compat(from->si_addr);
3322 #ifdef __ARCH_SI_TRAPNO
3323 		to->si_trapno = from->si_trapno;
3324 #endif
3325 		to->si_addr_lsb = from->si_addr_lsb;
3326 		break;
3327 	case SIL_FAULT_BNDERR:
3328 		to->si_addr = ptr_to_compat(from->si_addr);
3329 #ifdef __ARCH_SI_TRAPNO
3330 		to->si_trapno = from->si_trapno;
3331 #endif
3332 		to->si_lower = ptr_to_compat(from->si_lower);
3333 		to->si_upper = ptr_to_compat(from->si_upper);
3334 		break;
3335 	case SIL_FAULT_PKUERR:
3336 		to->si_addr = ptr_to_compat(from->si_addr);
3337 #ifdef __ARCH_SI_TRAPNO
3338 		to->si_trapno = from->si_trapno;
3339 #endif
3340 		to->si_pkey = from->si_pkey;
3341 		break;
3342 	case SIL_CHLD:
3343 		to->si_pid = from->si_pid;
3344 		to->si_uid = from->si_uid;
3345 		to->si_status = from->si_status;
3346 		to->si_utime = from->si_utime;
3347 		to->si_stime = from->si_stime;
3348 		break;
3349 	case SIL_RT:
3350 		to->si_pid = from->si_pid;
3351 		to->si_uid = from->si_uid;
3352 		to->si_int = from->si_int;
3353 		break;
3354 	case SIL_SYS:
3355 		to->si_call_addr = ptr_to_compat(from->si_call_addr);
3356 		to->si_syscall   = from->si_syscall;
3357 		to->si_arch      = from->si_arch;
3358 		break;
3359 	}
3360 }
3361 
3362 int __copy_siginfo_to_user32(struct compat_siginfo __user *to,
3363 			   const struct kernel_siginfo *from)
3364 {
3365 	struct compat_siginfo new;
3366 
3367 	copy_siginfo_to_external32(&new, from);
3368 	if (copy_to_user(to, &new, sizeof(struct compat_siginfo)))
3369 		return -EFAULT;
3370 	return 0;
3371 }
3372 
3373 static int post_copy_siginfo_from_user32(kernel_siginfo_t *to,
3374 					 const struct compat_siginfo *from)
3375 {
3376 	clear_siginfo(to);
3377 	to->si_signo = from->si_signo;
3378 	to->si_errno = from->si_errno;
3379 	to->si_code  = from->si_code;
3380 	switch(siginfo_layout(from->si_signo, from->si_code)) {
3381 	case SIL_KILL:
3382 		to->si_pid = from->si_pid;
3383 		to->si_uid = from->si_uid;
3384 		break;
3385 	case SIL_TIMER:
3386 		to->si_tid     = from->si_tid;
3387 		to->si_overrun = from->si_overrun;
3388 		to->si_int     = from->si_int;
3389 		break;
3390 	case SIL_POLL:
3391 		to->si_band = from->si_band;
3392 		to->si_fd   = from->si_fd;
3393 		break;
3394 	case SIL_FAULT:
3395 		to->si_addr = compat_ptr(from->si_addr);
3396 #ifdef __ARCH_SI_TRAPNO
3397 		to->si_trapno = from->si_trapno;
3398 #endif
3399 		break;
3400 	case SIL_FAULT_MCEERR:
3401 		to->si_addr = compat_ptr(from->si_addr);
3402 #ifdef __ARCH_SI_TRAPNO
3403 		to->si_trapno = from->si_trapno;
3404 #endif
3405 		to->si_addr_lsb = from->si_addr_lsb;
3406 		break;
3407 	case SIL_FAULT_BNDERR:
3408 		to->si_addr = compat_ptr(from->si_addr);
3409 #ifdef __ARCH_SI_TRAPNO
3410 		to->si_trapno = from->si_trapno;
3411 #endif
3412 		to->si_lower = compat_ptr(from->si_lower);
3413 		to->si_upper = compat_ptr(from->si_upper);
3414 		break;
3415 	case SIL_FAULT_PKUERR:
3416 		to->si_addr = compat_ptr(from->si_addr);
3417 #ifdef __ARCH_SI_TRAPNO
3418 		to->si_trapno = from->si_trapno;
3419 #endif
3420 		to->si_pkey = from->si_pkey;
3421 		break;
3422 	case SIL_CHLD:
3423 		to->si_pid    = from->si_pid;
3424 		to->si_uid    = from->si_uid;
3425 		to->si_status = from->si_status;
3426 #ifdef CONFIG_X86_X32_ABI
3427 		if (in_x32_syscall()) {
3428 			to->si_utime = from->_sifields._sigchld_x32._utime;
3429 			to->si_stime = from->_sifields._sigchld_x32._stime;
3430 		} else
3431 #endif
3432 		{
3433 			to->si_utime = from->si_utime;
3434 			to->si_stime = from->si_stime;
3435 		}
3436 		break;
3437 	case SIL_RT:
3438 		to->si_pid = from->si_pid;
3439 		to->si_uid = from->si_uid;
3440 		to->si_int = from->si_int;
3441 		break;
3442 	case SIL_SYS:
3443 		to->si_call_addr = compat_ptr(from->si_call_addr);
3444 		to->si_syscall   = from->si_syscall;
3445 		to->si_arch      = from->si_arch;
3446 		break;
3447 	}
3448 	return 0;
3449 }
3450 
3451 static int __copy_siginfo_from_user32(int signo, struct kernel_siginfo *to,
3452 				      const struct compat_siginfo __user *ufrom)
3453 {
3454 	struct compat_siginfo from;
3455 
3456 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3457 		return -EFAULT;
3458 
3459 	from.si_signo = signo;
3460 	return post_copy_siginfo_from_user32(to, &from);
3461 }
3462 
3463 int copy_siginfo_from_user32(struct kernel_siginfo *to,
3464 			     const struct compat_siginfo __user *ufrom)
3465 {
3466 	struct compat_siginfo from;
3467 
3468 	if (copy_from_user(&from, ufrom, sizeof(struct compat_siginfo)))
3469 		return -EFAULT;
3470 
3471 	return post_copy_siginfo_from_user32(to, &from);
3472 }
3473 #endif /* CONFIG_COMPAT */
3474 
3475 /**
3476  *  do_sigtimedwait - wait for queued signals specified in @which
3477  *  @which: queued signals to wait for
3478  *  @info: if non-null, the signal's siginfo is returned here
3479  *  @ts: upper bound on process time suspension
3480  */
3481 static int do_sigtimedwait(const sigset_t *which, kernel_siginfo_t *info,
3482 		    const struct timespec64 *ts)
3483 {
3484 	ktime_t *to = NULL, timeout = KTIME_MAX;
3485 	struct task_struct *tsk = current;
3486 	sigset_t mask = *which;
3487 	int sig, ret = 0;
3488 
3489 	if (ts) {
3490 		if (!timespec64_valid(ts))
3491 			return -EINVAL;
3492 		timeout = timespec64_to_ktime(*ts);
3493 		to = &timeout;
3494 	}
3495 
3496 	/*
3497 	 * Invert the set of allowed signals to get those we want to block.
3498 	 */
3499 	sigdelsetmask(&mask, sigmask(SIGKILL) | sigmask(SIGSTOP));
3500 	signotset(&mask);
3501 
3502 	spin_lock_irq(&tsk->sighand->siglock);
3503 	sig = dequeue_signal(tsk, &mask, info);
3504 	if (!sig && timeout) {
3505 		/*
3506 		 * None ready, temporarily unblock those we're interested
3507 		 * while we are sleeping in so that we'll be awakened when
3508 		 * they arrive. Unblocking is always fine, we can avoid
3509 		 * set_current_blocked().
3510 		 */
3511 		tsk->real_blocked = tsk->blocked;
3512 		sigandsets(&tsk->blocked, &tsk->blocked, &mask);
3513 		recalc_sigpending();
3514 		spin_unlock_irq(&tsk->sighand->siglock);
3515 
3516 		__set_current_state(TASK_INTERRUPTIBLE);
3517 		ret = freezable_schedule_hrtimeout_range(to, tsk->timer_slack_ns,
3518 							 HRTIMER_MODE_REL);
3519 		spin_lock_irq(&tsk->sighand->siglock);
3520 		__set_task_blocked(tsk, &tsk->real_blocked);
3521 		sigemptyset(&tsk->real_blocked);
3522 		sig = dequeue_signal(tsk, &mask, info);
3523 	}
3524 	spin_unlock_irq(&tsk->sighand->siglock);
3525 
3526 	if (sig)
3527 		return sig;
3528 	return ret ? -EINTR : -EAGAIN;
3529 }
3530 
3531 /**
3532  *  sys_rt_sigtimedwait - synchronously wait for queued signals specified
3533  *			in @uthese
3534  *  @uthese: queued signals to wait for
3535  *  @uinfo: if non-null, the signal's siginfo is returned here
3536  *  @uts: upper bound on process time suspension
3537  *  @sigsetsize: size of sigset_t type
3538  */
3539 SYSCALL_DEFINE4(rt_sigtimedwait, const sigset_t __user *, uthese,
3540 		siginfo_t __user *, uinfo,
3541 		const struct __kernel_timespec __user *, uts,
3542 		size_t, sigsetsize)
3543 {
3544 	sigset_t these;
3545 	struct timespec64 ts;
3546 	kernel_siginfo_t info;
3547 	int ret;
3548 
3549 	/* XXX: Don't preclude handling different sized sigset_t's.  */
3550 	if (sigsetsize != sizeof(sigset_t))
3551 		return -EINVAL;
3552 
3553 	if (copy_from_user(&these, uthese, sizeof(these)))
3554 		return -EFAULT;
3555 
3556 	if (uts) {
3557 		if (get_timespec64(&ts, uts))
3558 			return -EFAULT;
3559 	}
3560 
3561 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3562 
3563 	if (ret > 0 && uinfo) {
3564 		if (copy_siginfo_to_user(uinfo, &info))
3565 			ret = -EFAULT;
3566 	}
3567 
3568 	return ret;
3569 }
3570 
3571 #ifdef CONFIG_COMPAT_32BIT_TIME
3572 SYSCALL_DEFINE4(rt_sigtimedwait_time32, const sigset_t __user *, uthese,
3573 		siginfo_t __user *, uinfo,
3574 		const struct old_timespec32 __user *, uts,
3575 		size_t, sigsetsize)
3576 {
3577 	sigset_t these;
3578 	struct timespec64 ts;
3579 	kernel_siginfo_t info;
3580 	int ret;
3581 
3582 	if (sigsetsize != sizeof(sigset_t))
3583 		return -EINVAL;
3584 
3585 	if (copy_from_user(&these, uthese, sizeof(these)))
3586 		return -EFAULT;
3587 
3588 	if (uts) {
3589 		if (get_old_timespec32(&ts, uts))
3590 			return -EFAULT;
3591 	}
3592 
3593 	ret = do_sigtimedwait(&these, &info, uts ? &ts : NULL);
3594 
3595 	if (ret > 0 && uinfo) {
3596 		if (copy_siginfo_to_user(uinfo, &info))
3597 			ret = -EFAULT;
3598 	}
3599 
3600 	return ret;
3601 }
3602 #endif
3603 
3604 #ifdef CONFIG_COMPAT
3605 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time64, compat_sigset_t __user *, uthese,
3606 		struct compat_siginfo __user *, uinfo,
3607 		struct __kernel_timespec __user *, uts, compat_size_t, sigsetsize)
3608 {
3609 	sigset_t s;
3610 	struct timespec64 t;
3611 	kernel_siginfo_t info;
3612 	long ret;
3613 
3614 	if (sigsetsize != sizeof(sigset_t))
3615 		return -EINVAL;
3616 
3617 	if (get_compat_sigset(&s, uthese))
3618 		return -EFAULT;
3619 
3620 	if (uts) {
3621 		if (get_timespec64(&t, uts))
3622 			return -EFAULT;
3623 	}
3624 
3625 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3626 
3627 	if (ret > 0 && uinfo) {
3628 		if (copy_siginfo_to_user32(uinfo, &info))
3629 			ret = -EFAULT;
3630 	}
3631 
3632 	return ret;
3633 }
3634 
3635 #ifdef CONFIG_COMPAT_32BIT_TIME
3636 COMPAT_SYSCALL_DEFINE4(rt_sigtimedwait_time32, compat_sigset_t __user *, uthese,
3637 		struct compat_siginfo __user *, uinfo,
3638 		struct old_timespec32 __user *, uts, compat_size_t, sigsetsize)
3639 {
3640 	sigset_t s;
3641 	struct timespec64 t;
3642 	kernel_siginfo_t info;
3643 	long ret;
3644 
3645 	if (sigsetsize != sizeof(sigset_t))
3646 		return -EINVAL;
3647 
3648 	if (get_compat_sigset(&s, uthese))
3649 		return -EFAULT;
3650 
3651 	if (uts) {
3652 		if (get_old_timespec32(&t, uts))
3653 			return -EFAULT;
3654 	}
3655 
3656 	ret = do_sigtimedwait(&s, &info, uts ? &t : NULL);
3657 
3658 	if (ret > 0 && uinfo) {
3659 		if (copy_siginfo_to_user32(uinfo, &info))
3660 			ret = -EFAULT;
3661 	}
3662 
3663 	return ret;
3664 }
3665 #endif
3666 #endif
3667 
3668 static inline void prepare_kill_siginfo(int sig, struct kernel_siginfo *info)
3669 {
3670 	clear_siginfo(info);
3671 	info->si_signo = sig;
3672 	info->si_errno = 0;
3673 	info->si_code = SI_USER;
3674 	info->si_pid = task_tgid_vnr(current);
3675 	info->si_uid = from_kuid_munged(current_user_ns(), current_uid());
3676 }
3677 
3678 /**
3679  *  sys_kill - send a signal to a process
3680  *  @pid: the PID of the process
3681  *  @sig: signal to be sent
3682  */
3683 SYSCALL_DEFINE2(kill, pid_t, pid, int, sig)
3684 {
3685 	struct kernel_siginfo info;
3686 
3687 	prepare_kill_siginfo(sig, &info);
3688 
3689 	return kill_something_info(sig, &info, pid);
3690 }
3691 
3692 /*
3693  * Verify that the signaler and signalee either are in the same pid namespace
3694  * or that the signaler's pid namespace is an ancestor of the signalee's pid
3695  * namespace.
3696  */
3697 static bool access_pidfd_pidns(struct pid *pid)
3698 {
3699 	struct pid_namespace *active = task_active_pid_ns(current);
3700 	struct pid_namespace *p = ns_of_pid(pid);
3701 
3702 	for (;;) {
3703 		if (!p)
3704 			return false;
3705 		if (p == active)
3706 			break;
3707 		p = p->parent;
3708 	}
3709 
3710 	return true;
3711 }
3712 
3713 static int copy_siginfo_from_user_any(kernel_siginfo_t *kinfo,
3714 		siginfo_t __user *info)
3715 {
3716 #ifdef CONFIG_COMPAT
3717 	/*
3718 	 * Avoid hooking up compat syscalls and instead handle necessary
3719 	 * conversions here. Note, this is a stop-gap measure and should not be
3720 	 * considered a generic solution.
3721 	 */
3722 	if (in_compat_syscall())
3723 		return copy_siginfo_from_user32(
3724 			kinfo, (struct compat_siginfo __user *)info);
3725 #endif
3726 	return copy_siginfo_from_user(kinfo, info);
3727 }
3728 
3729 static struct pid *pidfd_to_pid(const struct file *file)
3730 {
3731 	struct pid *pid;
3732 
3733 	pid = pidfd_pid(file);
3734 	if (!IS_ERR(pid))
3735 		return pid;
3736 
3737 	return tgid_pidfd_to_pid(file);
3738 }
3739 
3740 /**
3741  * sys_pidfd_send_signal - Signal a process through a pidfd
3742  * @pidfd:  file descriptor of the process
3743  * @sig:    signal to send
3744  * @info:   signal info
3745  * @flags:  future flags
3746  *
3747  * The syscall currently only signals via PIDTYPE_PID which covers
3748  * kill(<positive-pid>, <signal>. It does not signal threads or process
3749  * groups.
3750  * In order to extend the syscall to threads and process groups the @flags
3751  * argument should be used. In essence, the @flags argument will determine
3752  * what is signaled and not the file descriptor itself. Put in other words,
3753  * grouping is a property of the flags argument not a property of the file
3754  * descriptor.
3755  *
3756  * Return: 0 on success, negative errno on failure
3757  */
3758 SYSCALL_DEFINE4(pidfd_send_signal, int, pidfd, int, sig,
3759 		siginfo_t __user *, info, unsigned int, flags)
3760 {
3761 	int ret;
3762 	struct fd f;
3763 	struct pid *pid;
3764 	kernel_siginfo_t kinfo;
3765 
3766 	/* Enforce flags be set to 0 until we add an extension. */
3767 	if (flags)
3768 		return -EINVAL;
3769 
3770 	f = fdget(pidfd);
3771 	if (!f.file)
3772 		return -EBADF;
3773 
3774 	/* Is this a pidfd? */
3775 	pid = pidfd_to_pid(f.file);
3776 	if (IS_ERR(pid)) {
3777 		ret = PTR_ERR(pid);
3778 		goto err;
3779 	}
3780 
3781 	ret = -EINVAL;
3782 	if (!access_pidfd_pidns(pid))
3783 		goto err;
3784 
3785 	if (info) {
3786 		ret = copy_siginfo_from_user_any(&kinfo, info);
3787 		if (unlikely(ret))
3788 			goto err;
3789 
3790 		ret = -EINVAL;
3791 		if (unlikely(sig != kinfo.si_signo))
3792 			goto err;
3793 
3794 		/* Only allow sending arbitrary signals to yourself. */
3795 		ret = -EPERM;
3796 		if ((task_pid(current) != pid) &&
3797 		    (kinfo.si_code >= 0 || kinfo.si_code == SI_TKILL))
3798 			goto err;
3799 	} else {
3800 		prepare_kill_siginfo(sig, &kinfo);
3801 	}
3802 
3803 	ret = kill_pid_info(sig, &kinfo, pid);
3804 
3805 err:
3806 	fdput(f);
3807 	return ret;
3808 }
3809 
3810 static int
3811 do_send_specific(pid_t tgid, pid_t pid, int sig, struct kernel_siginfo *info)
3812 {
3813 	struct task_struct *p;
3814 	int error = -ESRCH;
3815 
3816 	rcu_read_lock();
3817 	p = find_task_by_vpid(pid);
3818 	if (p && (tgid <= 0 || task_tgid_vnr(p) == tgid)) {
3819 		error = check_kill_permission(sig, info, p);
3820 		/*
3821 		 * The null signal is a permissions and process existence
3822 		 * probe.  No signal is actually delivered.
3823 		 */
3824 		if (!error && sig) {
3825 			error = do_send_sig_info(sig, info, p, PIDTYPE_PID);
3826 			/*
3827 			 * If lock_task_sighand() failed we pretend the task
3828 			 * dies after receiving the signal. The window is tiny,
3829 			 * and the signal is private anyway.
3830 			 */
3831 			if (unlikely(error == -ESRCH))
3832 				error = 0;
3833 		}
3834 	}
3835 	rcu_read_unlock();
3836 
3837 	return error;
3838 }
3839 
3840 static int do_tkill(pid_t tgid, pid_t pid, int sig)
3841 {
3842 	struct kernel_siginfo info;
3843 
3844 	clear_siginfo(&info);
3845 	info.si_signo = sig;
3846 	info.si_errno = 0;
3847 	info.si_code = SI_TKILL;
3848 	info.si_pid = task_tgid_vnr(current);
3849 	info.si_uid = from_kuid_munged(current_user_ns(), current_uid());
3850 
3851 	return do_send_specific(tgid, pid, sig, &info);
3852 }
3853 
3854 /**
3855  *  sys_tgkill - send signal to one specific thread
3856  *  @tgid: the thread group ID of the thread
3857  *  @pid: the PID of the thread
3858  *  @sig: signal to be sent
3859  *
3860  *  This syscall also checks the @tgid and returns -ESRCH even if the PID
3861  *  exists but it's not belonging to the target process anymore. This
3862  *  method solves the problem of threads exiting and PIDs getting reused.
3863  */
3864 SYSCALL_DEFINE3(tgkill, pid_t, tgid, pid_t, pid, int, sig)
3865 {
3866 	/* This is only valid for single tasks */
3867 	if (pid <= 0 || tgid <= 0)
3868 		return -EINVAL;
3869 
3870 	return do_tkill(tgid, pid, sig);
3871 }
3872 
3873 /**
3874  *  sys_tkill - send signal to one specific task
3875  *  @pid: the PID of the task
3876  *  @sig: signal to be sent
3877  *
3878  *  Send a signal to only one task, even if it's a CLONE_THREAD task.
3879  */
3880 SYSCALL_DEFINE2(tkill, pid_t, pid, int, sig)
3881 {
3882 	/* This is only valid for single tasks */
3883 	if (pid <= 0)
3884 		return -EINVAL;
3885 
3886 	return do_tkill(0, pid, sig);
3887 }
3888 
3889 static int do_rt_sigqueueinfo(pid_t pid, int sig, kernel_siginfo_t *info)
3890 {
3891 	/* Not even root can pretend to send signals from the kernel.
3892 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3893 	 */
3894 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3895 	    (task_pid_vnr(current) != pid))
3896 		return -EPERM;
3897 
3898 	/* POSIX.1b doesn't mention process groups.  */
3899 	return kill_proc_info(sig, info, pid);
3900 }
3901 
3902 /**
3903  *  sys_rt_sigqueueinfo - send signal information to a signal
3904  *  @pid: the PID of the thread
3905  *  @sig: signal to be sent
3906  *  @uinfo: signal info to be sent
3907  */
3908 SYSCALL_DEFINE3(rt_sigqueueinfo, pid_t, pid, int, sig,
3909 		siginfo_t __user *, uinfo)
3910 {
3911 	kernel_siginfo_t info;
3912 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
3913 	if (unlikely(ret))
3914 		return ret;
3915 	return do_rt_sigqueueinfo(pid, sig, &info);
3916 }
3917 
3918 #ifdef CONFIG_COMPAT
3919 COMPAT_SYSCALL_DEFINE3(rt_sigqueueinfo,
3920 			compat_pid_t, pid,
3921 			int, sig,
3922 			struct compat_siginfo __user *, uinfo)
3923 {
3924 	kernel_siginfo_t info;
3925 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
3926 	if (unlikely(ret))
3927 		return ret;
3928 	return do_rt_sigqueueinfo(pid, sig, &info);
3929 }
3930 #endif
3931 
3932 static int do_rt_tgsigqueueinfo(pid_t tgid, pid_t pid, int sig, kernel_siginfo_t *info)
3933 {
3934 	/* This is only valid for single tasks */
3935 	if (pid <= 0 || tgid <= 0)
3936 		return -EINVAL;
3937 
3938 	/* Not even root can pretend to send signals from the kernel.
3939 	 * Nor can they impersonate a kill()/tgkill(), which adds source info.
3940 	 */
3941 	if ((info->si_code >= 0 || info->si_code == SI_TKILL) &&
3942 	    (task_pid_vnr(current) != pid))
3943 		return -EPERM;
3944 
3945 	return do_send_specific(tgid, pid, sig, info);
3946 }
3947 
3948 SYSCALL_DEFINE4(rt_tgsigqueueinfo, pid_t, tgid, pid_t, pid, int, sig,
3949 		siginfo_t __user *, uinfo)
3950 {
3951 	kernel_siginfo_t info;
3952 	int ret = __copy_siginfo_from_user(sig, &info, uinfo);
3953 	if (unlikely(ret))
3954 		return ret;
3955 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3956 }
3957 
3958 #ifdef CONFIG_COMPAT
3959 COMPAT_SYSCALL_DEFINE4(rt_tgsigqueueinfo,
3960 			compat_pid_t, tgid,
3961 			compat_pid_t, pid,
3962 			int, sig,
3963 			struct compat_siginfo __user *, uinfo)
3964 {
3965 	kernel_siginfo_t info;
3966 	int ret = __copy_siginfo_from_user32(sig, &info, uinfo);
3967 	if (unlikely(ret))
3968 		return ret;
3969 	return do_rt_tgsigqueueinfo(tgid, pid, sig, &info);
3970 }
3971 #endif
3972 
3973 /*
3974  * For kthreads only, must not be used if cloned with CLONE_SIGHAND
3975  */
3976 void kernel_sigaction(int sig, __sighandler_t action)
3977 {
3978 	spin_lock_irq(&current->sighand->siglock);
3979 	current->sighand->action[sig - 1].sa.sa_handler = action;
3980 	if (action == SIG_IGN) {
3981 		sigset_t mask;
3982 
3983 		sigemptyset(&mask);
3984 		sigaddset(&mask, sig);
3985 
3986 		flush_sigqueue_mask(&mask, &current->signal->shared_pending);
3987 		flush_sigqueue_mask(&mask, &current->pending);
3988 		recalc_sigpending();
3989 	}
3990 	spin_unlock_irq(&current->sighand->siglock);
3991 }
3992 EXPORT_SYMBOL(kernel_sigaction);
3993 
3994 void __weak sigaction_compat_abi(struct k_sigaction *act,
3995 		struct k_sigaction *oact)
3996 {
3997 }
3998 
3999 int do_sigaction(int sig, struct k_sigaction *act, struct k_sigaction *oact)
4000 {
4001 	struct task_struct *p = current, *t;
4002 	struct k_sigaction *k;
4003 	sigset_t mask;
4004 
4005 	if (!valid_signal(sig) || sig < 1 || (act && sig_kernel_only(sig)))
4006 		return -EINVAL;
4007 
4008 	k = &p->sighand->action[sig-1];
4009 
4010 	spin_lock_irq(&p->sighand->siglock);
4011 	if (oact)
4012 		*oact = *k;
4013 
4014 	/*
4015 	 * Make sure that we never accidentally claim to support SA_UNSUPPORTED,
4016 	 * e.g. by having an architecture use the bit in their uapi.
4017 	 */
4018 	BUILD_BUG_ON(UAPI_SA_FLAGS & SA_UNSUPPORTED);
4019 
4020 	/*
4021 	 * Clear unknown flag bits in order to allow userspace to detect missing
4022 	 * support for flag bits and to allow the kernel to use non-uapi bits
4023 	 * internally.
4024 	 */
4025 	if (act)
4026 		act->sa.sa_flags &= UAPI_SA_FLAGS;
4027 	if (oact)
4028 		oact->sa.sa_flags &= UAPI_SA_FLAGS;
4029 
4030 	sigaction_compat_abi(act, oact);
4031 
4032 	if (act) {
4033 		sigdelsetmask(&act->sa.sa_mask,
4034 			      sigmask(SIGKILL) | sigmask(SIGSTOP));
4035 		*k = *act;
4036 		/*
4037 		 * POSIX 3.3.1.3:
4038 		 *  "Setting a signal action to SIG_IGN for a signal that is
4039 		 *   pending shall cause the pending signal to be discarded,
4040 		 *   whether or not it is blocked."
4041 		 *
4042 		 *  "Setting a signal action to SIG_DFL for a signal that is
4043 		 *   pending and whose default action is to ignore the signal
4044 		 *   (for example, SIGCHLD), shall cause the pending signal to
4045 		 *   be discarded, whether or not it is blocked"
4046 		 */
4047 		if (sig_handler_ignored(sig_handler(p, sig), sig)) {
4048 			sigemptyset(&mask);
4049 			sigaddset(&mask, sig);
4050 			flush_sigqueue_mask(&mask, &p->signal->shared_pending);
4051 			for_each_thread(p, t)
4052 				flush_sigqueue_mask(&mask, &t->pending);
4053 		}
4054 	}
4055 
4056 	spin_unlock_irq(&p->sighand->siglock);
4057 	return 0;
4058 }
4059 
4060 static int
4061 do_sigaltstack (const stack_t *ss, stack_t *oss, unsigned long sp,
4062 		size_t min_ss_size)
4063 {
4064 	struct task_struct *t = current;
4065 
4066 	if (oss) {
4067 		memset(oss, 0, sizeof(stack_t));
4068 		oss->ss_sp = (void __user *) t->sas_ss_sp;
4069 		oss->ss_size = t->sas_ss_size;
4070 		oss->ss_flags = sas_ss_flags(sp) |
4071 			(current->sas_ss_flags & SS_FLAG_BITS);
4072 	}
4073 
4074 	if (ss) {
4075 		void __user *ss_sp = ss->ss_sp;
4076 		size_t ss_size = ss->ss_size;
4077 		unsigned ss_flags = ss->ss_flags;
4078 		int ss_mode;
4079 
4080 		if (unlikely(on_sig_stack(sp)))
4081 			return -EPERM;
4082 
4083 		ss_mode = ss_flags & ~SS_FLAG_BITS;
4084 		if (unlikely(ss_mode != SS_DISABLE && ss_mode != SS_ONSTACK &&
4085 				ss_mode != 0))
4086 			return -EINVAL;
4087 
4088 		if (ss_mode == SS_DISABLE) {
4089 			ss_size = 0;
4090 			ss_sp = NULL;
4091 		} else {
4092 			if (unlikely(ss_size < min_ss_size))
4093 				return -ENOMEM;
4094 		}
4095 
4096 		t->sas_ss_sp = (unsigned long) ss_sp;
4097 		t->sas_ss_size = ss_size;
4098 		t->sas_ss_flags = ss_flags;
4099 	}
4100 	return 0;
4101 }
4102 
4103 SYSCALL_DEFINE2(sigaltstack,const stack_t __user *,uss, stack_t __user *,uoss)
4104 {
4105 	stack_t new, old;
4106 	int err;
4107 	if (uss && copy_from_user(&new, uss, sizeof(stack_t)))
4108 		return -EFAULT;
4109 	err = do_sigaltstack(uss ? &new : NULL, uoss ? &old : NULL,
4110 			      current_user_stack_pointer(),
4111 			      MINSIGSTKSZ);
4112 	if (!err && uoss && copy_to_user(uoss, &old, sizeof(stack_t)))
4113 		err = -EFAULT;
4114 	return err;
4115 }
4116 
4117 int restore_altstack(const stack_t __user *uss)
4118 {
4119 	stack_t new;
4120 	if (copy_from_user(&new, uss, sizeof(stack_t)))
4121 		return -EFAULT;
4122 	(void)do_sigaltstack(&new, NULL, current_user_stack_pointer(),
4123 			     MINSIGSTKSZ);
4124 	/* squash all but EFAULT for now */
4125 	return 0;
4126 }
4127 
4128 int __save_altstack(stack_t __user *uss, unsigned long sp)
4129 {
4130 	struct task_struct *t = current;
4131 	int err = __put_user((void __user *)t->sas_ss_sp, &uss->ss_sp) |
4132 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4133 		__put_user(t->sas_ss_size, &uss->ss_size);
4134 	if (err)
4135 		return err;
4136 	if (t->sas_ss_flags & SS_AUTODISARM)
4137 		sas_ss_reset(t);
4138 	return 0;
4139 }
4140 
4141 #ifdef CONFIG_COMPAT
4142 static int do_compat_sigaltstack(const compat_stack_t __user *uss_ptr,
4143 				 compat_stack_t __user *uoss_ptr)
4144 {
4145 	stack_t uss, uoss;
4146 	int ret;
4147 
4148 	if (uss_ptr) {
4149 		compat_stack_t uss32;
4150 		if (copy_from_user(&uss32, uss_ptr, sizeof(compat_stack_t)))
4151 			return -EFAULT;
4152 		uss.ss_sp = compat_ptr(uss32.ss_sp);
4153 		uss.ss_flags = uss32.ss_flags;
4154 		uss.ss_size = uss32.ss_size;
4155 	}
4156 	ret = do_sigaltstack(uss_ptr ? &uss : NULL, &uoss,
4157 			     compat_user_stack_pointer(),
4158 			     COMPAT_MINSIGSTKSZ);
4159 	if (ret >= 0 && uoss_ptr)  {
4160 		compat_stack_t old;
4161 		memset(&old, 0, sizeof(old));
4162 		old.ss_sp = ptr_to_compat(uoss.ss_sp);
4163 		old.ss_flags = uoss.ss_flags;
4164 		old.ss_size = uoss.ss_size;
4165 		if (copy_to_user(uoss_ptr, &old, sizeof(compat_stack_t)))
4166 			ret = -EFAULT;
4167 	}
4168 	return ret;
4169 }
4170 
4171 COMPAT_SYSCALL_DEFINE2(sigaltstack,
4172 			const compat_stack_t __user *, uss_ptr,
4173 			compat_stack_t __user *, uoss_ptr)
4174 {
4175 	return do_compat_sigaltstack(uss_ptr, uoss_ptr);
4176 }
4177 
4178 int compat_restore_altstack(const compat_stack_t __user *uss)
4179 {
4180 	int err = do_compat_sigaltstack(uss, NULL);
4181 	/* squash all but -EFAULT for now */
4182 	return err == -EFAULT ? err : 0;
4183 }
4184 
4185 int __compat_save_altstack(compat_stack_t __user *uss, unsigned long sp)
4186 {
4187 	int err;
4188 	struct task_struct *t = current;
4189 	err = __put_user(ptr_to_compat((void __user *)t->sas_ss_sp),
4190 			 &uss->ss_sp) |
4191 		__put_user(t->sas_ss_flags, &uss->ss_flags) |
4192 		__put_user(t->sas_ss_size, &uss->ss_size);
4193 	if (err)
4194 		return err;
4195 	if (t->sas_ss_flags & SS_AUTODISARM)
4196 		sas_ss_reset(t);
4197 	return 0;
4198 }
4199 #endif
4200 
4201 #ifdef __ARCH_WANT_SYS_SIGPENDING
4202 
4203 /**
4204  *  sys_sigpending - examine pending signals
4205  *  @uset: where mask of pending signal is returned
4206  */
4207 SYSCALL_DEFINE1(sigpending, old_sigset_t __user *, uset)
4208 {
4209 	sigset_t set;
4210 
4211 	if (sizeof(old_sigset_t) > sizeof(*uset))
4212 		return -EINVAL;
4213 
4214 	do_sigpending(&set);
4215 
4216 	if (copy_to_user(uset, &set, sizeof(old_sigset_t)))
4217 		return -EFAULT;
4218 
4219 	return 0;
4220 }
4221 
4222 #ifdef CONFIG_COMPAT
4223 COMPAT_SYSCALL_DEFINE1(sigpending, compat_old_sigset_t __user *, set32)
4224 {
4225 	sigset_t set;
4226 
4227 	do_sigpending(&set);
4228 
4229 	return put_user(set.sig[0], set32);
4230 }
4231 #endif
4232 
4233 #endif
4234 
4235 #ifdef __ARCH_WANT_SYS_SIGPROCMASK
4236 /**
4237  *  sys_sigprocmask - examine and change blocked signals
4238  *  @how: whether to add, remove, or set signals
4239  *  @nset: signals to add or remove (if non-null)
4240  *  @oset: previous value of signal mask if non-null
4241  *
4242  * Some platforms have their own version with special arguments;
4243  * others support only sys_rt_sigprocmask.
4244  */
4245 
4246 SYSCALL_DEFINE3(sigprocmask, int, how, old_sigset_t __user *, nset,
4247 		old_sigset_t __user *, oset)
4248 {
4249 	old_sigset_t old_set, new_set;
4250 	sigset_t new_blocked;
4251 
4252 	old_set = current->blocked.sig[0];
4253 
4254 	if (nset) {
4255 		if (copy_from_user(&new_set, nset, sizeof(*nset)))
4256 			return -EFAULT;
4257 
4258 		new_blocked = current->blocked;
4259 
4260 		switch (how) {
4261 		case SIG_BLOCK:
4262 			sigaddsetmask(&new_blocked, new_set);
4263 			break;
4264 		case SIG_UNBLOCK:
4265 			sigdelsetmask(&new_blocked, new_set);
4266 			break;
4267 		case SIG_SETMASK:
4268 			new_blocked.sig[0] = new_set;
4269 			break;
4270 		default:
4271 			return -EINVAL;
4272 		}
4273 
4274 		set_current_blocked(&new_blocked);
4275 	}
4276 
4277 	if (oset) {
4278 		if (copy_to_user(oset, &old_set, sizeof(*oset)))
4279 			return -EFAULT;
4280 	}
4281 
4282 	return 0;
4283 }
4284 #endif /* __ARCH_WANT_SYS_SIGPROCMASK */
4285 
4286 #ifndef CONFIG_ODD_RT_SIGACTION
4287 /**
4288  *  sys_rt_sigaction - alter an action taken by a process
4289  *  @sig: signal to be sent
4290  *  @act: new sigaction
4291  *  @oact: used to save the previous sigaction
4292  *  @sigsetsize: size of sigset_t type
4293  */
4294 SYSCALL_DEFINE4(rt_sigaction, int, sig,
4295 		const struct sigaction __user *, act,
4296 		struct sigaction __user *, oact,
4297 		size_t, sigsetsize)
4298 {
4299 	struct k_sigaction new_sa, old_sa;
4300 	int ret;
4301 
4302 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4303 	if (sigsetsize != sizeof(sigset_t))
4304 		return -EINVAL;
4305 
4306 	if (act && copy_from_user(&new_sa.sa, act, sizeof(new_sa.sa)))
4307 		return -EFAULT;
4308 
4309 	ret = do_sigaction(sig, act ? &new_sa : NULL, oact ? &old_sa : NULL);
4310 	if (ret)
4311 		return ret;
4312 
4313 	if (oact && copy_to_user(oact, &old_sa.sa, sizeof(old_sa.sa)))
4314 		return -EFAULT;
4315 
4316 	return 0;
4317 }
4318 #ifdef CONFIG_COMPAT
4319 COMPAT_SYSCALL_DEFINE4(rt_sigaction, int, sig,
4320 		const struct compat_sigaction __user *, act,
4321 		struct compat_sigaction __user *, oact,
4322 		compat_size_t, sigsetsize)
4323 {
4324 	struct k_sigaction new_ka, old_ka;
4325 #ifdef __ARCH_HAS_SA_RESTORER
4326 	compat_uptr_t restorer;
4327 #endif
4328 	int ret;
4329 
4330 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4331 	if (sigsetsize != sizeof(compat_sigset_t))
4332 		return -EINVAL;
4333 
4334 	if (act) {
4335 		compat_uptr_t handler;
4336 		ret = get_user(handler, &act->sa_handler);
4337 		new_ka.sa.sa_handler = compat_ptr(handler);
4338 #ifdef __ARCH_HAS_SA_RESTORER
4339 		ret |= get_user(restorer, &act->sa_restorer);
4340 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4341 #endif
4342 		ret |= get_compat_sigset(&new_ka.sa.sa_mask, &act->sa_mask);
4343 		ret |= get_user(new_ka.sa.sa_flags, &act->sa_flags);
4344 		if (ret)
4345 			return -EFAULT;
4346 	}
4347 
4348 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4349 	if (!ret && oact) {
4350 		ret = put_user(ptr_to_compat(old_ka.sa.sa_handler),
4351 			       &oact->sa_handler);
4352 		ret |= put_compat_sigset(&oact->sa_mask, &old_ka.sa.sa_mask,
4353 					 sizeof(oact->sa_mask));
4354 		ret |= put_user(old_ka.sa.sa_flags, &oact->sa_flags);
4355 #ifdef __ARCH_HAS_SA_RESTORER
4356 		ret |= put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4357 				&oact->sa_restorer);
4358 #endif
4359 	}
4360 	return ret;
4361 }
4362 #endif
4363 #endif /* !CONFIG_ODD_RT_SIGACTION */
4364 
4365 #ifdef CONFIG_OLD_SIGACTION
4366 SYSCALL_DEFINE3(sigaction, int, sig,
4367 		const struct old_sigaction __user *, act,
4368 	        struct old_sigaction __user *, oact)
4369 {
4370 	struct k_sigaction new_ka, old_ka;
4371 	int ret;
4372 
4373 	if (act) {
4374 		old_sigset_t mask;
4375 		if (!access_ok(act, sizeof(*act)) ||
4376 		    __get_user(new_ka.sa.sa_handler, &act->sa_handler) ||
4377 		    __get_user(new_ka.sa.sa_restorer, &act->sa_restorer) ||
4378 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4379 		    __get_user(mask, &act->sa_mask))
4380 			return -EFAULT;
4381 #ifdef __ARCH_HAS_KA_RESTORER
4382 		new_ka.ka_restorer = NULL;
4383 #endif
4384 		siginitset(&new_ka.sa.sa_mask, mask);
4385 	}
4386 
4387 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4388 
4389 	if (!ret && oact) {
4390 		if (!access_ok(oact, sizeof(*oact)) ||
4391 		    __put_user(old_ka.sa.sa_handler, &oact->sa_handler) ||
4392 		    __put_user(old_ka.sa.sa_restorer, &oact->sa_restorer) ||
4393 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4394 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4395 			return -EFAULT;
4396 	}
4397 
4398 	return ret;
4399 }
4400 #endif
4401 #ifdef CONFIG_COMPAT_OLD_SIGACTION
4402 COMPAT_SYSCALL_DEFINE3(sigaction, int, sig,
4403 		const struct compat_old_sigaction __user *, act,
4404 	        struct compat_old_sigaction __user *, oact)
4405 {
4406 	struct k_sigaction new_ka, old_ka;
4407 	int ret;
4408 	compat_old_sigset_t mask;
4409 	compat_uptr_t handler, restorer;
4410 
4411 	if (act) {
4412 		if (!access_ok(act, sizeof(*act)) ||
4413 		    __get_user(handler, &act->sa_handler) ||
4414 		    __get_user(restorer, &act->sa_restorer) ||
4415 		    __get_user(new_ka.sa.sa_flags, &act->sa_flags) ||
4416 		    __get_user(mask, &act->sa_mask))
4417 			return -EFAULT;
4418 
4419 #ifdef __ARCH_HAS_KA_RESTORER
4420 		new_ka.ka_restorer = NULL;
4421 #endif
4422 		new_ka.sa.sa_handler = compat_ptr(handler);
4423 		new_ka.sa.sa_restorer = compat_ptr(restorer);
4424 		siginitset(&new_ka.sa.sa_mask, mask);
4425 	}
4426 
4427 	ret = do_sigaction(sig, act ? &new_ka : NULL, oact ? &old_ka : NULL);
4428 
4429 	if (!ret && oact) {
4430 		if (!access_ok(oact, sizeof(*oact)) ||
4431 		    __put_user(ptr_to_compat(old_ka.sa.sa_handler),
4432 			       &oact->sa_handler) ||
4433 		    __put_user(ptr_to_compat(old_ka.sa.sa_restorer),
4434 			       &oact->sa_restorer) ||
4435 		    __put_user(old_ka.sa.sa_flags, &oact->sa_flags) ||
4436 		    __put_user(old_ka.sa.sa_mask.sig[0], &oact->sa_mask))
4437 			return -EFAULT;
4438 	}
4439 	return ret;
4440 }
4441 #endif
4442 
4443 #ifdef CONFIG_SGETMASK_SYSCALL
4444 
4445 /*
4446  * For backwards compatibility.  Functionality superseded by sigprocmask.
4447  */
4448 SYSCALL_DEFINE0(sgetmask)
4449 {
4450 	/* SMP safe */
4451 	return current->blocked.sig[0];
4452 }
4453 
4454 SYSCALL_DEFINE1(ssetmask, int, newmask)
4455 {
4456 	int old = current->blocked.sig[0];
4457 	sigset_t newset;
4458 
4459 	siginitset(&newset, newmask);
4460 	set_current_blocked(&newset);
4461 
4462 	return old;
4463 }
4464 #endif /* CONFIG_SGETMASK_SYSCALL */
4465 
4466 #ifdef __ARCH_WANT_SYS_SIGNAL
4467 /*
4468  * For backwards compatibility.  Functionality superseded by sigaction.
4469  */
4470 SYSCALL_DEFINE2(signal, int, sig, __sighandler_t, handler)
4471 {
4472 	struct k_sigaction new_sa, old_sa;
4473 	int ret;
4474 
4475 	new_sa.sa.sa_handler = handler;
4476 	new_sa.sa.sa_flags = SA_ONESHOT | SA_NOMASK;
4477 	sigemptyset(&new_sa.sa.sa_mask);
4478 
4479 	ret = do_sigaction(sig, &new_sa, &old_sa);
4480 
4481 	return ret ? ret : (unsigned long)old_sa.sa.sa_handler;
4482 }
4483 #endif /* __ARCH_WANT_SYS_SIGNAL */
4484 
4485 #ifdef __ARCH_WANT_SYS_PAUSE
4486 
4487 SYSCALL_DEFINE0(pause)
4488 {
4489 	while (!signal_pending(current)) {
4490 		__set_current_state(TASK_INTERRUPTIBLE);
4491 		schedule();
4492 	}
4493 	return -ERESTARTNOHAND;
4494 }
4495 
4496 #endif
4497 
4498 static int sigsuspend(sigset_t *set)
4499 {
4500 	current->saved_sigmask = current->blocked;
4501 	set_current_blocked(set);
4502 
4503 	while (!signal_pending(current)) {
4504 		__set_current_state(TASK_INTERRUPTIBLE);
4505 		schedule();
4506 	}
4507 	set_restore_sigmask();
4508 	return -ERESTARTNOHAND;
4509 }
4510 
4511 /**
4512  *  sys_rt_sigsuspend - replace the signal mask for a value with the
4513  *	@unewset value until a signal is received
4514  *  @unewset: new signal mask value
4515  *  @sigsetsize: size of sigset_t type
4516  */
4517 SYSCALL_DEFINE2(rt_sigsuspend, sigset_t __user *, unewset, size_t, sigsetsize)
4518 {
4519 	sigset_t newset;
4520 
4521 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4522 	if (sigsetsize != sizeof(sigset_t))
4523 		return -EINVAL;
4524 
4525 	if (copy_from_user(&newset, unewset, sizeof(newset)))
4526 		return -EFAULT;
4527 	return sigsuspend(&newset);
4528 }
4529 
4530 #ifdef CONFIG_COMPAT
4531 COMPAT_SYSCALL_DEFINE2(rt_sigsuspend, compat_sigset_t __user *, unewset, compat_size_t, sigsetsize)
4532 {
4533 	sigset_t newset;
4534 
4535 	/* XXX: Don't preclude handling different sized sigset_t's.  */
4536 	if (sigsetsize != sizeof(sigset_t))
4537 		return -EINVAL;
4538 
4539 	if (get_compat_sigset(&newset, unewset))
4540 		return -EFAULT;
4541 	return sigsuspend(&newset);
4542 }
4543 #endif
4544 
4545 #ifdef CONFIG_OLD_SIGSUSPEND
4546 SYSCALL_DEFINE1(sigsuspend, old_sigset_t, mask)
4547 {
4548 	sigset_t blocked;
4549 	siginitset(&blocked, mask);
4550 	return sigsuspend(&blocked);
4551 }
4552 #endif
4553 #ifdef CONFIG_OLD_SIGSUSPEND3
4554 SYSCALL_DEFINE3(sigsuspend, int, unused1, int, unused2, old_sigset_t, mask)
4555 {
4556 	sigset_t blocked;
4557 	siginitset(&blocked, mask);
4558 	return sigsuspend(&blocked);
4559 }
4560 #endif
4561 
4562 __weak const char *arch_vma_name(struct vm_area_struct *vma)
4563 {
4564 	return NULL;
4565 }
4566 
4567 static inline void siginfo_buildtime_checks(void)
4568 {
4569 	BUILD_BUG_ON(sizeof(struct siginfo) != SI_MAX_SIZE);
4570 
4571 	/* Verify the offsets in the two siginfos match */
4572 #define CHECK_OFFSET(field) \
4573 	BUILD_BUG_ON(offsetof(siginfo_t, field) != offsetof(kernel_siginfo_t, field))
4574 
4575 	/* kill */
4576 	CHECK_OFFSET(si_pid);
4577 	CHECK_OFFSET(si_uid);
4578 
4579 	/* timer */
4580 	CHECK_OFFSET(si_tid);
4581 	CHECK_OFFSET(si_overrun);
4582 	CHECK_OFFSET(si_value);
4583 
4584 	/* rt */
4585 	CHECK_OFFSET(si_pid);
4586 	CHECK_OFFSET(si_uid);
4587 	CHECK_OFFSET(si_value);
4588 
4589 	/* sigchld */
4590 	CHECK_OFFSET(si_pid);
4591 	CHECK_OFFSET(si_uid);
4592 	CHECK_OFFSET(si_status);
4593 	CHECK_OFFSET(si_utime);
4594 	CHECK_OFFSET(si_stime);
4595 
4596 	/* sigfault */
4597 	CHECK_OFFSET(si_addr);
4598 	CHECK_OFFSET(si_addr_lsb);
4599 	CHECK_OFFSET(si_lower);
4600 	CHECK_OFFSET(si_upper);
4601 	CHECK_OFFSET(si_pkey);
4602 
4603 	/* sigpoll */
4604 	CHECK_OFFSET(si_band);
4605 	CHECK_OFFSET(si_fd);
4606 
4607 	/* sigsys */
4608 	CHECK_OFFSET(si_call_addr);
4609 	CHECK_OFFSET(si_syscall);
4610 	CHECK_OFFSET(si_arch);
4611 #undef CHECK_OFFSET
4612 
4613 	/* usb asyncio */
4614 	BUILD_BUG_ON(offsetof(struct siginfo, si_pid) !=
4615 		     offsetof(struct siginfo, si_addr));
4616 	if (sizeof(int) == sizeof(void __user *)) {
4617 		BUILD_BUG_ON(sizeof_field(struct siginfo, si_pid) !=
4618 			     sizeof(void __user *));
4619 	} else {
4620 		BUILD_BUG_ON((sizeof_field(struct siginfo, si_pid) +
4621 			      sizeof_field(struct siginfo, si_uid)) !=
4622 			     sizeof(void __user *));
4623 		BUILD_BUG_ON(offsetofend(struct siginfo, si_pid) !=
4624 			     offsetof(struct siginfo, si_uid));
4625 	}
4626 #ifdef CONFIG_COMPAT
4627 	BUILD_BUG_ON(offsetof(struct compat_siginfo, si_pid) !=
4628 		     offsetof(struct compat_siginfo, si_addr));
4629 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4630 		     sizeof(compat_uptr_t));
4631 	BUILD_BUG_ON(sizeof_field(struct compat_siginfo, si_pid) !=
4632 		     sizeof_field(struct siginfo, si_pid));
4633 #endif
4634 }
4635 
4636 void __init signals_init(void)
4637 {
4638 	siginfo_buildtime_checks();
4639 
4640 	sigqueue_cachep = KMEM_CACHE(sigqueue, SLAB_PANIC);
4641 }
4642 
4643 #ifdef CONFIG_KGDB_KDB
4644 #include <linux/kdb.h>
4645 /*
4646  * kdb_send_sig - Allows kdb to send signals without exposing
4647  * signal internals.  This function checks if the required locks are
4648  * available before calling the main signal code, to avoid kdb
4649  * deadlocks.
4650  */
4651 void kdb_send_sig(struct task_struct *t, int sig)
4652 {
4653 	static struct task_struct *kdb_prev_t;
4654 	int new_t, ret;
4655 	if (!spin_trylock(&t->sighand->siglock)) {
4656 		kdb_printf("Can't do kill command now.\n"
4657 			   "The sigmask lock is held somewhere else in "
4658 			   "kernel, try again later\n");
4659 		return;
4660 	}
4661 	new_t = kdb_prev_t != t;
4662 	kdb_prev_t = t;
4663 	if (t->state != TASK_RUNNING && new_t) {
4664 		spin_unlock(&t->sighand->siglock);
4665 		kdb_printf("Process is not RUNNING, sending a signal from "
4666 			   "kdb risks deadlock\n"
4667 			   "on the run queue locks. "
4668 			   "The signal has _not_ been sent.\n"
4669 			   "Reissue the kill command if you want to risk "
4670 			   "the deadlock.\n");
4671 		return;
4672 	}
4673 	ret = send_signal(sig, SEND_SIG_PRIV, t, PIDTYPE_PID);
4674 	spin_unlock(&t->sighand->siglock);
4675 	if (ret)
4676 		kdb_printf("Fail to deliver Signal %d to process %d.\n",
4677 			   sig, t->pid);
4678 	else
4679 		kdb_printf("Signal %d is sent to process %d.\n", sig, t->pid);
4680 }
4681 #endif	/* CONFIG_KGDB_KDB */
4682